Environmental Impact of Invertebrates For Biological Control 0f Anthropods: Methods and Risk Assessment

Environmental Impact of Invertebrates for Biological Control of Arthropods Methods and Risk Assessment This page inte...

Author: Franz Bigler | Dirk Babendreier | Ulrich Kuhlmann

12 downloads 638 Views 2MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Environmental Impact of Invertebrates for Biological Control of Arthropods

Methods and Risk Assessment

This page intentionally left blank

Environmental Impact of Invertebrates for Biological Control of Arthropods Methods and Risk Assessment

Edited by

Franz Bigler and Dirk Babendreier Agroscope, FAL Reckenholz Swiss Federal Research Station for Agroecology and Agriculture Zürich Switzerland and

Ulrich Kuhlmann CABI Bioscience Switzerland Centre Delémont Switzerland

CABI Publishing

CABI Publishing is a division of CAB International CABI Publishing CAB International Wallingford Oxon OX10 8DE UK

CABI Publishing 875 Massachusetts Avenue 7th Floor Cambridge, MA 02139 USA

Tel: +44 (0)1491 832111 Fax: +44 (0)1491 833508 E-mail: [email protected] Website: www.cabi-publishing.org

Tel: +1 617 395 4056 Fax: +1 617 354 6875 E-mail: [email protected]

© CAB International 2006. All rights reserved. No part of this publication may be reproduced in any form or by any means, electronically, mechanically, by photocopying, recording or otherwise, without the prior permission of the copyright owners. A catalogue record for this book is available from the British Library, London, UK. Library of Congress Cataloging-in-Publication Data Environment impact of invertebrates for biological control of arthropods : methods and risk assessment / edited by Franz Bigler and Dirk Babendreier and Ulli Kuhlmann. p. cm. Includes bibliographical references and index. ISBN-13: 978-0-85199-058-3 (alk. paper) ISBN-10: 0-85199-058-4 (alk. paper) 1. Insect pests--Biological control. 2. Arthropod pests--Biological control. 3. Arthropoda as biological pest control agents. 4. Pesticides-Environmental aspects. I. Bigler, Franz. II. Babendreier, Dirk. III. Kuhlmann, Ulli. IV. Title. SB933.3.E58 2006 632⬘.96--dc22 2005020627 ISBN-10: ISBN-13:

0-85199-058-4 978-0-85199-058-3

Typeset by Columns Design Ltd, Reading, UK. Printed and bound in the UK by Cromwell Press, Trowbridge.

Contents

Contributors Foreword Joop C. van Lenteren Preface Acknowledgements

vii xi xiii xv

1

Current Status and Constraints in the Assessment of Non-target Effects Dirk Babendreier, Franz Bigler and Ulrich Kuhlmann

1

2

Selection of Non-target Species for Host Specificity Testing Ulrich Kuhlmann, Urs Schaffner and Peter G. Mason

15

3

Host Specificity in Arthropod Biological Control, Methods for Testing and Interpretation of the Data Joop C. van Lenteren, Matthew J.W. Cock, Thomas S. Hoffmeister and Don P.A. Sands

38

4

Measuring and Predicting Indirect Impacts of Biological Control: Competition, Displacement and Secondary Interactions Russell Messing, Bernard Roitberg and Jacques Brodeur

64

5

Risks of Interbreeding Between Species Used in Biological Control and Native Species, and Methods for Evaluating Their Occurrence and Impact Keith R. Hopper, Seth C. Britch and Eric Wajnberg

78

6

Assessing the Establishment Potential of Inundative Biological Control Agents Guy Boivin, Ursula M. Kölliker-Ott, Jeffrey Bale and Franz Bigler

98

7

Methods for Monitoring the Dispersal of Natural Enemies from Point Source Releases Associated with Augmentative Biological Control Nick J. Mills, Dirk Babendreier and Antoon J.M. Loomans

114

8

Risks of Plant Damage Caused by Natural Enemies Introduced for Arthropod Biological Control Ramon Albajes, Cristina Castañé, Rosa Gabarra and Òscar Alomar

132

v

vi

9

Contents

Methods for Assessment of Contaminants of Invertebrate Biological Control Agents and Associated Risks Mark S. Goettel and G. Douglas Inglis

145

10 Post-release Evaluation of Non-target Effects of Biological Control Agents Barbara I.P. Barratt, Bernd Blossey, Heikki M.T. Hokkanen

166

11 Molecular Methods for the Identification of Biological Control Agents at the Species and Strain Level Richard Stouthamer

187

12 The Usefulness of the Ecoregion Concept for Safer Import of Invertebrate Biological Control Agents Matthew J.W. Cock, Ulrich Kuhlmann, Urs Schaffner, Franz Bigler and Dirk Babendreier

202

13 Statistical Tools to Improve the Quality of Experiments and Data Analysis for Assessing Non-target Effects Thomas S. Hoffmeister, Dirk Babendreier and Eric Wajnberg

222

14 Principles of Environmental Risk Assessment with Emphasis on the New Zealand Perspective Abdul Moeed, Robert Hickson and Barbara I.P. Barratt

241

15 Environmental Risk Assessment: Methods for Comprehensive Evaluation and Quick Scan Joop C. van Lenteren and Antoon J.M. Loomans

254

16 Balancing Environmental Risks and Benefits: a Basic Approach Franz Bigler and Ursula M. Kölliker-Ott

273

Glossary

287

Index

291

Contributors

Albajes, Ramon, Universitat de Lleida, Centre UdL-IRTA, Rovira Roure 191, 25198 Lleida, Spain. Email: [email protected]. Phone number: ⫹34-973-702571. Fax number: ⫹34-973-238301. Alomar, Òscar, IRTA, Centre de Cabrils, 08348 Cabrils (Barcelona), Spain. Email: [email protected]. Phone number: ⫹34-93-750-9961. Fax number: ⫹34-93-7533954. Babendreier, Dirk, Agroscope, FAL Reckenholz, Swiss Federal Research Station for Agroecology and Agriculture, Reckenholzstr. 191, 8046 Zürich, Switzerland. Email: [email protected]. Phone number: ⫹41-44-377-7217. Fax number: ⫹4144-377-7201. Bale, Jeffrey, School of Biosciences, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK. Email: [email protected]. Phone number: ⫹44-121-414-5908. Fax number: ⫹44-121-414-5925. Barratt, Barbara, AgResearch, Invermay Agricultural Centre, Private Bag 50034, Mosgiel, New Zealand. Email: [email protected]. Phone number: ⫹64-3-489-9059. Fax number: ⫹64-3-489-3739. Bigler, Franz, Agroscope, FAL Reckenholz, Swiss Federal Research Station for Agroecology and Agriculture, Reckenholzstr. 191, 8046 Zürich, Switzerland. Email: [email protected]. Phone number: ⫹41-44-377-7235. Fax number: ⫹41-44377-7201. Blossey, Bernd, Department of Natural Resources, 122E Fernow Hall, Cornell University, Ithaca, New York 14853, USA. Email: [email protected]. Phone number: ⫹1-607-2555314. Fax number: ⫹1-607-255-0349. Boivin, Guy, Centre de Recherche et de Développement en Horticulture, Agriculture et Agroalimentaire Canada, 430 Boul. Gouin, Saint-Jean-sur-Richelieu, Québec J3B 3E6, Canada. Email: [email protected]. Phone number: ⫹1-450-346-4494. Fax number: ⫹1-450-346-7740. Britch, Seth, Beneficial Insects Introduction Research Laboratory, Agricultural Research Service, USDA, 501 South Chapel Street, Newark, DE 19713, USA. Email: [email protected]. Phone number: ⫹1-302-731-7330 ext. 239. Fax number: ⫹1-302-737-6780. Brodeur, Jacques, Département des Sciences Biologiques, Institut de Recherche en Biologie Végétale, Université de Montréal, 4101, rue Sherbrooke Est, Montréal vii

viii

Contributors

(Québec), Canada H1X 2B2. Email: [email protected]. Phone number: ⫹1-514-872-4563. Fax number: ⫹1-514-872-9406. Castañé, Cristina, IRTA, Centre de Cabrils, 08348 Cabrils, (Barcelona), Spain. Email: [email protected]. Phone number: ⫹34-93-750-9961. Fax number: ⫹34-93-7533954. Cock, Matthew, CABI Bioscience Switzerland Centre, Rue des Grillons 1, 2800 Delémont, Switzerland. Email: [email protected]. Phone number: ⫹41-32-421-4870. Fax number: ⫹41-32-421-4871. Gabarra, Rosa, IRTA, Centre de Cabrils, 08348 Cabrils (Barcelona), Spain. Email: [email protected]. Phone number: ⫹34-93-750-9976. Fax number: ⫹34-93-7533954. Goettel, Mark, Environmental Health, Agriculture and Agri-Food Canada, Lethbridge Research Centre, PO Box 3000, 5403 – 1st Avenue South, Lethbridge, AB T1J 4B1 Canada. Email: [email protected]. Phone number: ⫹44-403-317-2264. Fax number: ⫹44-403-382-3156. Hickson, Robert, Ministry of Research, Science and Technology, PO Box 5336, Wellington, New Zealand. Email: [email protected]. Phone number: ⫹644-917-2917. Fax number: ⫹64-4-471-1284. Hoffmeister, Thomas, Institute of Ecology and Evolutionary Biology, University of Bremen, Leobener Str. NW2, D-28359 Bremen, Germany. Email: [email protected]. Phone number: ⫹49-421-218-4290. Fax number: ⫹49-421-218-4504. Hokkanen, Heikki, Department of Applied Zoology, University of Helsinki, PO Box 27, 00014 Helsinki, Finland. Email: heikki.hokkanen@helsinki.fi. Phone number: ⫹3589191-58371. Fax number: ⫹358-9191-58463. Hopper, Keith, Beneficial Insects Introduction Research Laboratory, Agricultural Research Service, USDA, 501 South Chapel Street, Newark, DE 19713, USA. Email: [email protected]. Phone number: ⫹1-302-731-7330 ext. 238. Fax number: ⫹1-302737-6780. Inglis, Douglas, Food Safety and Quality, Agriculture and Agri-Food Canada, 5403-1st Avenue South, Lethbridge, AB T1J 4B1, Canada. Email: [email protected]. Phone number: ⫹1-403-317-3355. Fax number: ⫹1-403-382-3156. Kölliker-Ott, Ursula, Agroscope, FAL Reckenholz, Swiss Federal Research Station for Agroecology and Agriculture, Reckenholzstr. 191, 8046 Zürich, Switzerland. Email: [email protected]. Phone number: ⫹41-44-377-7181. Fax number: ⫹4144-377-7201. Kuhlmann, Ulli, CABI Bioscience Switzerland Centre, Rue des Grillons 1, 2800 Delémont, Switzerland. Email: [email protected]. Phone number: ⫹41-32-4214882. Fax number: ⫹41-32-421-4871. Loomans, Antoon, Plant Protection Service, Section Entomology, PO Box 9102, 6700 HC Wageningen, The Netherlands. Email: [email protected]. Phone number: ⫹31-317-496825. Fax number: ⫹31-317-421701. Mason, Peter, Agriculture and Agri-food Canada, Research Centre, K.W. Neatby Building, Central Experimental Farm, 960, Carling Avenue, Ottawa, Ontario K1A OC6, Canada. Email: [email protected]. Phone number: ⫹1-613-759-1908. Fax number: ⫹1-613-759-170. Messing, Russell, University of Hawaii at Manoa, Kauai Agricultural Research Center, 7370 Kuamoo Road, Kapaa, Hawaii 96746, USA. Email: [email protected]. Phone number: ⫹1-808-822-4984 x223. Fax number: ⫹1-808-822-2190. Mills, Nick, Environmental Science, Policy and Management, 127 Mulford Hall, University of California, Berkeley, CA 94720-3114, USA. Email: [email protected]. Phone number: ⫹1-510-642-1711. Fax number: ⫹1-510-643-5438.

Contributors

ix

Moeed, Abdul, ERMA New Zealand, PO Box 131, Wellington, New Zealand. Email: [email protected]. Phone number: ⫹64-4-916-2426. Fax number: ⫹64-4914-0433. Roitberg, Bernie, Department of Biological Science, Simon Fraser University, Burnaby, BC, V5A IS6, Canada. Email: [email protected]. Phone number: ⫹1-604-2913585. Fax number: ⫹1-604-291-3496. Sands, Don, CSIRO Entomology, 120 Meiers Road, Indooroopilly, Queensland 4068, Australia. Email: [email protected]. Phone number: ⫹61-403-517224. Schaffner, Urs, CABI Bioscience Switzerland Centre, Rue des Grillons 1, 2800 Delémont, Switzerland. Email: [email protected]. Phone number: ⫹41-32-4214877. Fax number: ⫹41-32-421-4871. Stouthamer, Richard, Department of Entomology, University of California, Riverside, CA 92521, USA. Email: [email protected]. Phone number: ⫹1-951-8272422. Fax number: ⫹1-951-827-3086. van Lenteren, Joop, Laboratory of Entomology, Wageningen University, PO Box 8031, 6700 EH Wageningen, The Netherlands. Email: [email protected]. Phone number: ⫹31-317-482327. Fax number: ⫹31-317-484821. Wajnberg, Eric, INRA, 400, Route des Chappes, BP 167, 06903 Sophia Antipolis Cedex, France. Email: [email protected]. Phone number: ⫹33-4-92-38-6447. Fax number: ⫹33-4-92-38-6557.


Foreword

Classical biological control of insects, where exotic natural enemies are introduced to control exotic pests, has been applied for more than 120 years, and release of more than 2000 species of natural enemies has resulted in the permanent reduction of at least 165 pest species worldwide. Augmentative biological control, where exotic or native natural enemies are periodically released, has been used for 90 years, and more than 150 species of natural enemy are available on demand for the control of about 100 pest species. Contrary to the thorough environmental risk evaluations applied in the search for natural enemies of weeds, potential risks of biological control agents for arthropod control have not been routinely studied in pre-release evaluations. The reason might be that until now, very few problems have been reported concerning negative effects of releases of invertebrates for control of arthropods, despite there having been well over 5000 introductions that have been made worldwide. It is a well-known fact that intended or accidental invasions by many other exotic organisms have resulted in serious negative environmental and economic effects. However, discussion of the risks of releases of exotic natural enemies for non-target species now takes a prominent place in biological control programmes. On the other hand, one normally tends to forget or even not know the enormous economic and environmental benefits resulting from biological control with introduced exotic organisms. Recent retrospective analyses of biological control projects have provided quantitative data on nontarget effects and illustrated the need for risk assessments to increase the future safety of biological control. Twenty countries have already implemented regulation for release of biological control agents and many other countries are considering regulation. Soon, the International Standard for Phytosanitary Measures (ISPM3) will become the standard for all biological control introductions worldwide, but this standard does not provide methods by which to assess environmental risks. The same can be said about other risk assessments that have previously been used to evaluate exotic natural enemies. In order to fulfil the need of developing environmental risk assessment methods, as well as a framework for a general risk assessment of biological control agents, an international group of scientists first wrote a number of working papers. Next, these were discussed and modified during a week of hard work in the Swiss mountains. Finally, the papers were peer reviewed and rewritten for the current book. The goal of this book is not only to present risk assessment methods, but also to give ample background information relevant for developing and adapting these methods. xi

xii

Foreword

It is my hope that this book will find its way to scientists, biological control workers and regulators. Intensive collaboration between representatives of these groups will hopefully result in a light and harmonized regulation procedure that is not prohibitive to the biological control industry and will result in the selection of safe natural enemies. Joop C. van Lenteren President of the International Organization for Biological Control (IOBC Global) Professor of Entomology, Laboratory of Entomology, Wageningen University, The Netherlands.

Preface

While safety of biological control was generally not questioned until the beginning of the 1990s, an ongoing debate started shortly after the Rio Convention on Biodiversity was agreed in 1992. Based on this agreement and on an increasing amount of published literature blaming biological control for contributing to biodiversity loss, international organizations and national governments started after the mid-nineties to publish documents in which general principles of guidance and good governance for import and release of invertebrate biological control agents were laid down. None of the international documents was meant to give detailed advice to national regulatory bodies on how to regulate import and release of such organisms, nor did they provide methodologies on how to assess potential effects and how to perform risk and benefit analysis. While the documents specify what information will be needed for risk assessment, they do not give any indication on how to obtain the relevant information, i.e., what methods could best be applied to obtain the needed data to perform risk assessments. This lack of background information and advice on methodology was the starting point of the present book. The idea was born to publish a document that summarizes the present status on risk assessment in biological control of arthropods with invertebrates, and gives guidance on methods to generate data which enable biological control scientists, natural enemy producers, retailers, practitioners and regulators to make informed risk assessments. The guiding principle of the book is to provide a science-based framework for identifying and evaluating relevant environmental effects that could result from import and release of exotic invertebrate biological control agents. It should assist those who are involved in risk and benefit assessment and in regulation of invertebrate biological control agents used against arthropod pests. A literature review has shown us that there is presently very little literature published providing reliable information on standard methods that could be applied to produce data for risk assessment. It is the intention of the present book to set a framework for risk assessment, to discuss strengths, weaknesses and lack of methods and to propose new approaches and practical guidance on how to measure and evaluate effects that contribute to environmental risks and benefits. We are aware that we do not cover all relevant aspects of risk–benefit assessment and regulation, and further efforts will be needed. Nevertheless, we are confident we can offer the reader a range of methods and guidance that will improve and facilitate regulation of invertebrate biological control agents, and contribute to the ongoing debate. xiii

xiv

Preface

Based on previous projects and existing experience on risk assessment and regulation of invertebrate biological control agents, we identified the most critical issues to be considered and addressed in this book. With financial support from the Swiss Agency of Environment, Landscape and Forest and the Swiss Federal Research Station for Agroecology and Agriculture we were able to invite an international group of experts to prepare chapters and to present and discuss them at a workshop held in Engelberg, Switzerland, in 2004. The very open, critical and constructive atmosphere here was the ground for fruitful debates that contributed to improving the chapters and to make the book more comprehensive. The book consists of three parts, namely the major section in which methods for assessing environmental effects of invertebrate biological control agents are reviewed, discussed in the light of risk assessment and, when possible, recommendations on appropriate methods are made. The second section consists of three chapters presenting different technical tools which are extremely important in environmental risk assessment and regulatory procedures, and they belong to the basic prerequisites to evaluate risks. In the third section, the principles of environmental risk assessment are presented together with a case study; two methods on how to perform risk analysis with invertebrate biological control agents are shown with practical examples given, and finally, a risk–benefit assessment together with an example is discussed. As the book is a compilation of the current knowledge of methodology available for assessing non-target effects and risks of invertebrate biological control agents, it shows the arsenal of tools and methods. However, limitations of our understanding of ecological mechanisms and lack of methods to analyse such processes show the obvious gaps. We are far from having answers and solutions to all questions relevant to risks and regulation, and we still need to tackle a number of practical problems. Bearing in mind that improvements can still be made in the future, we should not forget that regulation of biological control agents must be cost effective. Overregulation of biological control would be disastrous because it would prevent progress of biological control and its role in IPM. Regulation of invertebrate biological control agents will certainly undergo changes in the coming years. We expect that national authorities in many countries will be more demanding, with the consequent need for biological control manufacturers to prepare more elaborated dossiers, with more information and data. This will be an additional burden for biological control projects and lead to a longer time period for approval of new organisms. On the other hand, it will give more confidence in biological control and help to maintain and strengthen the good reputation of these pest control methods. We have reached our goals if this book contributes to the better assessment of environmental effects, risks and benefits of invertebrate biological control agents, and if it provides guidance to all those who are involved in biological control and its regulation. Franz Bigler, Dirk Babendreier and Ulli Kuhlmann, June 2005 Zürich and Delémont, Switzerland.

Acknowledgements

This book has been written by authors who have long-standing expertise in biological control and/or regulation of agents introduced and released to this end. First, we would like to thank those authors who participated in the workshop held in 2004 in Engelberg, in the Swiss Alps, where first drafts of the chapters were discussed and critically reviewed in an open and constructive spirit. Special thanks are addressed to the few coauthors who were not able to attend, but still made their invaluable contributions to different chapters. Many colleagues reviewed the chapters and gave their comments and views, provided ideas and insights and helped the authors to achieve a text which will be useful to all stakeholders of biological control. From within the group of workshop participants, we would like to thank Barbara Barratt, Guy Boivin, Jaques Brodeur, Keith Hopper, Doug Inglis, Antoon Loomans, Peter Mason, Russell Messing, Nick Mills, Bernie Roitberg, Richard Stouthamer and Joop van Lenteren. Furthermore, several external reviewers shared their expertise with us, and the following colleagues are particularly acknowledged: Moshe Coll, Eric Conti, Dave Gillespie, George Heimpel, Lia Hemerik, Mark Hoddle, Kim Hoelmer, Larry Lacey, Peter McEvoy, Bill Turnock, Franco Widmer and Robert Wiedenmann. This book is the fruit of a project funded by the Swiss Agency of Environment, Landscape and Forest, the Swiss Federal Research Station for Agroecology and Agriculture and CABI Bioscience Centre, Switzerland. We are thankful for the continuous support by these institutions.

xv


1

Current Status and Constraints in the Assessment of Non-target Effects Dirk Babendreier,1 Franz Bigler1 and Ulrich Kuhlmann2

1Agroscope,

FAL Reckenholz, Swiss Federal Research Station for Agroecology and Agriculture, Reckenholzstr. 191, 8046 Zürich, Switzerland (email: [email protected]; [email protected]; fax number: +41-44-3777201); 2CABI Bioscience Centre, Rue des Grillons 1, 2800 Delémont, Switzerland (email: [email protected]; fax number: +41-32-4214871)

Abstract In the last two decades increasing concerns have been expressed regarding potential nontarget effects of invertebrate biological control agents of arthropods. This has led to an increasing number of studies investigating non-target effects in many systems. Several international initiatives aimed at providing guidance for risk assessment of biological control agents are briefly reviewed here. Furthermore, we aim to provide an overview of the current status of non-target testing of arthropod biological control agents, and identify the most recent developments. Most importantly, we aim to identify constraints and unsolved questions which need further research or consideration in the future. Major obstacles encountered include the need for harmonization of regulation and methods, and the increasing costs that are associated with implementing regulation. In addition, statistical analysis, the interpretation of host range tests, and inherent uncertainties associated with non-target testing are major problems currently faced in risk assessment. Finally, this chapter will refer to other chapters of this book that address the identified issues and propose the urgently needed and relevant methodology.

History of Initiatives for Regulation The potential for non-target effects resulting from the release of biological control agents has been recognized for over a hundred years. However, only much later has this question stimulated intensive discussion among scientists and beyond (Howarth, 1983, 1991). Since then, nontarget effects in biological control are increasingly being studied, and a number

of reviews have been published within the last ten years (e.g. Simberloff and Stiling, 1996; Follett et al., 2000; Lockwood et al., 2001; Lynch et al., 2001; Louda et al., 2003). International laws and agreements coupled with an increasing interest in the import and release of exotic biological control agents requires harmonized and appropriate regulation. However, provisions within such legislation vary considerably

©CAB International 2006. Environmental Impact of Invertebrates for Biological Control of Arthropods: Methods and Risk Assessment (eds F. Bigler et al.)

1

2

D. Babendreier et al.

between countries. A starting point towards international regulation was marked by the FAO Code of Conduct for the Import and Release of Exotic Biological Control Agents; this was adopted in 1995 by the FAO Conference and published in 1996 as the International Standard for Phytosanitary Measures No. 3 (IPPC, 1996). One objective of the Code was to provide a standard for those countries that lack adequate legislation and procedures to regulate importation and to analyse risks related to biological control agents. The document lists in a generic way the responsibilities of the authorities and importers and exporters of biological control agents. The revised version of this Code of Conduct has extended its range from classical biological control to inundative biological control, native natural enemies, microorganisms and other beneficial organisms, and it also includes evaluation of environmental impacts (IPPC, 2005). This standard will certainly continue to provide guidance for countries that are developing their own legislative systems for biological control regulation, and the Code may be seen as a first attempt to globally harmonize regulation of biological control agents. Shortly after the Code’s first publication, the European and Mediterranean Plant Protection Organization (EPPO) together with CABI Bioscience organized a workshop on safety and efficacy of biological control in Europe (EPPO, 1997). This workshop broadly endorsed the FAO Code but recommended that regulation should not slow the importation or import of biological control agents, be it for preliminary research or for subsequent release. The workshop concluded that a certification system should be put in place for Europe instead of a registration procedure to ensure a ‘light’ regulatory system with efficient and rapid mechanisms. The reasoning behind this decision was based on previous experience with the registration system for microbial biological control agents in Europe: the EU Directive and its implementation is so stringent that it is basically impossible to register a new

microorganism in EU countries. An expert panel was established and the results of their meetings were published in two guidance documents and in a ‘positive list’ of organisms for safe use in EPPO countries (EPPO, 1999, 2001, 2002). The two guidelines stress the importance of a two-step system for importation and release, i.e. EU countries should first establish a regulatory process for the import of exotic organisms for research under containment. The results of these investigations will provide the necessary data to make decisions on whether the organism can later be imported for release. In parallel with the EPPO panel activities, the EU-funded research project ERBIC (Evaluating Environmental Risks of Biological Control Introductions into Europe) was executed from 1998 to 2002. One of the main outcomes of the project was a proposal for the environmental risk assessment of exotic natural enemies in inundative biological control (van Lenteren et al., 2003). This represents the first paper with detailed criteria for risk assessment and a ranking system that is based on the quantitative evaluation of more than 30 invertebrate biological control agents used in inundative control in Europe. In 2000, the North American Plant Protection Organization (NAPPO) published its ‘Guidelines for Petition for Release of Exotic Entomophagous Agents for the Biological Control of Pests’ (RSPM No 12, NAPPO, 2000). These guidelines are intended to assist researchers and companies in drafting a petition for release of exotic entomophagous agents for biological control of pest insects and mites. It will also assist reviewers and regulators in assessing the risk of exotic introductions intended for biological control. The guideline specifies the requirement for information on biology of the agent and the target pest(s), the economic impact of the pest, regulatory status, and the quarantine procedures needed for importation of the biological control agent. To some extent there has been some harmonization in data requirements for entomophagous biological control agents in that the three countries

Current Status and Constraints in the Assessment of Non-target Effects

(Canada, USA and Mexico) have agreed to conform to NAPPO guidelines. However, currently the regulatory system within the USA is cumbersome with a mixture of inconsistent Federal and State jurisdiction. The system for biological control regulation in Hawaii, the State where the most rigorous review procedure has been adopted, is worth reviewing. While the system appears to be exhaustive in ensuring environmental safety of biological control, and allows for a degree of public consultation, it is steeped in bureaucracy that results in frustration and lengthy delays for biological control practitioners. Island nations, such as Australia and New Zealand, have the unique situation where shared borders are not an issue, and complete control over imported biological control agents can be achieved. The 1996 Hazardous Substances and New Organisms (HSNO) Act in New Zealand (http://www.legislation.govt.nz) has attracted considerable attention internationally as very environmentally focused legislation, and its implementation by ERMA NZ has been observed with interest (see Moeed et al., Chapter 14 this volume). In Australia, biological control agents are regulated by two agencies under three separate Acts, and have been similarly heralded as a thorough and biosafety-conscious approach. The two systems have some key differences in approach, the most notable ones being the opportunity for public participation and the degree of risk-aversion of the regulatory agencies. An initiative starting from a meeting held in Canada in 1999 resulted in OECD (Organization for Economic Co-operation and Development) member countries developing a harmonized approach for regulation of invertebrate biological control agents. It was agreed that a harmonized regulatory system in the OECD member countries would be beneficial for biological control and that a ‘light’ form of regulation would be appropriate. The development of harmonized guidance for regulation requirements would enable companies to submit the same applications to many countries, and would allow regulatory agencies to benefit

3

from each other’s reviews. The document (OECD, 2003) proposes guidance for member countries on information requirements for: a) the characterization and identification of the organism; b) the assessment of safety and effects on human health; c) the assessment of environmental risks; and d) the assessment of efficacy of the organism. With native or established organisms and with those in long-term use in a country, substantially reduced information requirements may be appropriate. In Europe, the biological control industry expressed their concerns when the OECD guidance document was published as the information requirements were considered to be too stringent. As a consequence, the International Biocontrol Manufacturers’ Association (IBMA) proposed to the International Organization for Biological Control (IOBC/WPRS) facilitation of the harmonization among the European regulatory authorities. A Commission for the IOBC/WPRS was established in 2003 and a meeting of scientists, together with the biological control industry and regulators, resulted in the production of a document that gives detailed guidance on regulation procedures for exotic and indigenous biological control agents (Bigler et al., 2005). Most recently, the European Commission released a call for project applications with the aim of developing a balanced system for regulation of biological control agents (micro- and macro-organisms), semiochemicals and botanicals. This specifies that the number of microbiological products on the market in Europe is currently still low compared to other countries, e.g. the USA and Canada. The aim of the task is to review current legislation, guidelines and guidance documents and to compare this with similar legislation in other countries where the introduction of new biopesticides has proved to be more successful. New appropriate and balanced regulatory systems should be designed. It can be expected that within a few years the EU members and other European countries may regulate invertebrate biological control agents under uniform principles.

4


From this overview on regulation in different countries it is becoming evident that challenges and opportunities have emerged. The above-mentioned initiatives generally highlight what should be done or what knowledge is required, but they are not designed to provide detailed methods on how one should test for non-target effects. Recently, a guide to best practice of host range testing has been released by Van Driesche and Reardon (2004). In addition, all aspects of non-target testing have recently been addressed in a comprehensive review of the current methods used to assess potential risks of biological control agents (Babendreier et al., 2005). This book attempts to go a step further by providing guidance on methods necessary to assess non-target effects of invertebrate biological control agents of insect pests. The authors feel that the lack of methodology and approaches is a major concern and a bottleneck in environmental risk assessment at the moment, and that these issues need to be tackled.

Status and Important Issues in Assessing Environmental Effects While all documents underline the need for regulation of invertebrate biological control agents, the level of guidance on information needed for risk assessment varies to a great extent between these documents. The OECD guidance document (OECD, 2003) is one of the most comprehensive initiatives to date, as it requires relatively detailed information from the applicant in order to receive an import and release permit, and because the OECD covers a wide geographic area. Based on experience with many other regulatory documents released by the OECD, we assume that this document will be widely adopted internationally, or at least serve as a basis for national regulatory documents. Therefore, this chapter basically follows the issues raised in that document (OECD, 2003). While the first two parts of the document address issues of characterization and identification of organisms as well as

human health and safety, here we will discuss mainly the third part, i.e. the assessment of environmental risks.

Host specificity Host specificity is a key element if nontarget effects of biological control agents are to be assessed, and this is also reflected in the OECD document. Although only information available to identify any potential hazards posed to the environment is currently required under 3.1, data may be required for host specificity testing under 3.2 (Table 1.1). Here, we like to stress that host range assessment does not necessarily mean that tests have to be conducted. Often, published information is sufficient to draw conclusions on the host specificity of the agent. A recent example was provided by De Nardo and Hopper (2004), who conducted a comprehensive literature study for the ichneumonid parasitoid Macrocentrus grandii (Goidanich). These authors stressed that a lot of information can be obtained even from negative observations, i.e. from studies on potential nontarget hosts that did not report the biological control candidate as a natural enemy. Although host specificity testing has been required in weed biological control projects for many decades, it was incorporated into arthropod biological control projects rather recently. For the latter, there are still not many experimental studies available in which host range testing was conducted, though this number increased recently (see Babendreier et al., 2005). There are also several reviews or discussion papers available dealing with topics that need to be addressed in these tests (Sands, 1997, 1998; Van Driesche and Hoddle, 1997; Sands and Van Driesche, 2000; Van Driesche and Murray, 2004a,b; Van Driesche and Reardon, 2004). After the first step, i.e. the collection of all available information (Sands and Van Driesche, 2004) an important subsequent step may be to carry out field surveys in the


5

Table 1.1. Information requirements of the OECD document on environmental risk assessment of biological control agents. 3. Information for assessment of environmental risks 3.1 Identify any potential hazards posed to the environment including: (a) available information on the role of organism in original ecosystem, the type of natural enemy (parasitoid, predator, pathogen), type of organisms it attacks, effects of attack on targets and non-targets, intra-guild effects, higher up trophic level effects, effects on ecosystem (b) available information on existing natural enemies of the target organism in the area of release (c) available information on non-target effects from previous use in biological control 3.2 Host specificity testing (a) available information and/or data on possible direct effects: ● on non-target host/prey related to target host (phylogenetically or ecologically related) ● on non-related non-target hosts, such as threatened and endangered species ● concerning competition or displacement of organisms ● concerning potential for interbreeding with indigenous natural enemy strains or biotypes ● on plants (target crop and non-target plants) (b) available information and/or data on potential of establishment and dispersal of biological control agent (c) available information on and/or data on possible indirect effects (d) available information (from rearing facility; in the field) on ability to vector viruses or microorganisms which can negatively affect non-target organisms 3.3 Available information, and/or data on potential host/prey range in areas of release and potential distribution 3.4 Available information on environmental benefits e.g. beneficial effects of release compared to current or alternative control methods 3.5 Summary of information for assessment of environmental risks

country of origin and also to analyse the fauna of the proposed area of introduction (Hoddle, 2004). For those surveys, classical ecological methods or more recently developed molecular methods may be used depending on the organisms (Symondson, 2002; Gariepy et al., 2005). Field surveys are not only an important preliminary step in identifying the species with the most narrow host range out of a pool of species, but they can also provide guidance regarding which species should be included in host specificity tests (see Kuhlmann et al., Chapter 2, this volume). A general problem with field surveys is in defining the limits of the system. Should one collect only species from taxa that contain known hosts or include additional taxa? Creating a list of species that should be tested for acceptance by biological control agents is obviously a difficult task. A general problem, especially for insect biological control, is that the taxonomy of involved groups is often unclear (Van

Driesche and Reardon, 2004). Moreover, the number of species in taxonomic groups is often higher by an order of magnitude compared to plants. Molecular tools are increasingly being used and may help to solve this problem in the future (see Stouthamer, Chapter 11, this volume). Criteria that have been taken into account for creating such lists in arthropod biological control have included geographic distributions, oviposition phenology, number of generations per year, overwintering stage, host-plant preferences, and the type and feeding niche of the host (for a review, see Babendreier et al., 2005). In addition to the ecological criteria mentioned above, the importance and availability of potential non-target species were also considered; some species that would be desirable members of a host range test list may be impossible to find or to rear. However, there appears to be some contradiction as the OECD (2003, see Table 1.1) requires information on rare non-target hosts which, generally, is very difficult or

6


impossible to obtain (Barratt, 2004). In this book, Kuhlmann et al. (Chapter 2, this volume) for the first time worked out a general approach that could be applied in creating a list of non-target species used in host-range testing, both for inundative and classical biological control agents targeting insects. The ultimate aim of host range tests is to determine the agent’s ecological host range, i.e. the number of hosts that will be attacked in the field where the biological control agent is to be introduced (Van Driesche and Reardon, 2004). Clearly, laboratory tests have their limitations, as it is extremely difficult to accurately reproduce the cues and stimuli that affect host acceptance of biological control agents in a natural environment (Keller, 1999; Kuhlmann et al., 2000; Sands and Van Driesche, 2000). The interpretation of host specificity tests is a problem and there are ongoing debates regarding how indicative these tests are. A number of studies exist that conducted nochoice tests and choice tests with the same non-target species. The majority of these studies have shown that results from both kinds of tests are in general agreement (Duan and Messing, 2000; Zilahi-Balogh et al., 2002; Mansfield and Mills, 2004). However, Haye (2004) has shown that several non-target species were less preferred in choice tests while target and non-target species were equally parasitized in nochoice tests. Unfortunately, the reverse was also observed, i.e. non-targets and the target were similarly attacked in choice tests while less non-target parasitism was observed in the no-choice test. Whether choice tests are useful or necessary at all is still debated. Guidance on what test should be used and how this should be done is given by van Lenteren et al. (Chapter 3, this volume). Most importantly, however, one likes to know whether results obtained under laboratory (or semi-field) conditions are indicative of what a biological control agent would attack in the field. So far, there is no long track record on the reliability of host specificity testing in arthropod biological control. A pioneering study was conducted by Barratt et al. (1997), who compared results on host specificity of Microctonus

aethiopoides Loan and Microctonus hyperodae Loan (Hymenoptera: Braconidae) obtained in the laboratory with actual field parasitism after the agents were established. The authors basically concluded that tests conducted in the laboratory were in fact indicative of field parasitism. Coombs (2003) reported that the tachinid fly Trichopoda giacomelli (Blanchard) attacked two nontarget hosts after field release in Australia, exactly as was anticipated by host range tests carried out beforehand. However, there are also examples, such as the retrospective case study on the braconid wasp Peristenus digoneutis Loan (Haye et al., 2005), suggesting that physiological host range is often (much) greater than ecological host range. Despite the fact that laboratory tests demonstrated high parasitism levels in non-targets, ecological assessments in both North America and Europe suggested a much lower impact of P. digoneutis on non-target mirids. While some non-targets were not parasitized at all, others showed very low levels of parasitism (below 1% in Europe). Therefore, ecological host range studies in the area of origin provide useful supplementary data for interpretation of pre-release laboratory host range tests. Recently, Withers and Browne (2004) came up with a different approach, aiming the overall objective at maximizing the probability that non-target test species would be accepted during laboratory tests, which resulted in an accurate (although probably overestimated) risk assessment of the invertebrate biological control agent. When relying only on small cage laboratory experiments to assess the maximum host range possible retrospectively, P. digoneutis may have been classified as potentially risky, when in fact laboratory tests may have had a poor predictive value in this case. In general, when and why there is a good match between laboratory and field data remains an open and important question in arthropod biological control.

Competition and indirect effects It is suggested that negative interactions amongst biological control agents and com-


petitors may play a significant role both for the success of biological control projects and for non-target effects (Denoth et al., 2002; Reitz and Trumble, 2002). In fact, some well-documented examples of displacement have occurred among introduced biological control agents, and some of these showed that ecological processes responsible for displacement can be very complex (e.g. Murdoch et al., 1996). Regarding the natural enemy complex of the target, it is obvious that a successful biological control agent by itself may have dramatic consequences on the composition of this complex (e.g. Neuenschwander, 2001). It may be questioned whether displacement of an exotic natural enemy by another exotic, and population changes of native natural enemies associated with the control of the pest, can really be considered relevant non-target effects. Information on indirect effects is now required by the OECD (2003, see Table 1.1), but how this can be achieved is unclear, and it is still debated how an indirect effect can be defined. We believe that a clarifying definition has been provided by Messing et al. (Chapter 4, this volume), which basically distinguishes between direct competitive effects (those in which a natural enemy comes into direct physical contact with a competitor) and indirect competitive effects (in which the interaction among competing natural enemies is mediated via a third organism). While the former part of the definition relates to intra-guild predation (also listed in the OECD document, see Table 1.1), the latter part of the definition relates to all other, sometimes complex, processes. Messing et al. (Chapter 4, this volume) propose that an evaluation of indirect effects should preferentially concentrate on population- and community-level impacts rather than on consequences on individuals, and where possible, should be pursued under field conditions for extended periods of time. These studies typically include prolonged post-release monitoring and are thus labour-intensive and costly. Basic methods used to date include field surveys to compare non-target populations prior to and following release of the biological control agents (Brown, 2003), field

7

cage studies including the biological control agents and a competitor (Schellhorn et al., 2002) or intra-guild experiments in small arenas (e.g. for predators (Burgio et al., 2002); for parasitoids (Wang and Messing, 2002)). A tiered approach, combining laboratory, semi-field and field experiments, was recently applied in order to assess whether mass releases of Trichogramma brassicae Bezdenko (Hymenoptera: Trichogrammatidae) against the European corn borer might have detrimental effects on populations of other natural enemies in maize and adjacent habitats (Babendreier et al., 2003a). Again, the most serious problem with indirect effects is their complexity and the high degree of uncertainty inherently associated with them. Therefore, it is very difficult to incorporate them into risk assessment schemes (see Messing et al., Chapter 4 this volume; van Lenteren and Loomans, Chapter 15, this volume).

Post-release studies Retrospective post-release studies could be especially useful in verifying predictions made from host specificity testing before release of a biological control agent; however, to our knowledge, very few such studies are available (see Barratt et al., Chapter 10, this volume). This is probably due to the fact that most host specificity tests of arthropod biological control agents have been conducted only recently. The paucity of baseline data is a major drawback for postrelease studies. Typically for such studies, one or several non-target species were selected and sampled in areas where the biological control agent was released or was known to occur, and the mortality due to the agent was determined. Another method involves the placement of non-target individuals in the field where the biological control agent is known to occur or has been experimentally released. Life tables have also been shown to be a valuable tool in post-release studies, making it possible to put the observed mortality of the non-targets into context. To date, these studies suggest

8


that it is often not the suspected introduced agent, but rather other factors, that were responsible for most of the non-target mortality observed (Barron et al., 2003; Johnson et al., 2005). Alternatively, populations of non-targets could be observed both in areas where the biological control agent is present and in areas where it is absent. Although this approach is being used in New Zealand to detect potential non-target effects in longterm studies (B.I.P. Barratt, Mosgiel, NZ, 2004, personal communication), to date no published reports are available where this method has been applied in arthropod biological control.

Establishment and dispersal Two additional topics, namely the potential for establishment and dispersal, need to be addressed in risk assessment of biological control agents, though they are mainly important in inundative release programmes. Unfortunately, until recently (Babendreier et al., 2005) there have been very few published studies dealing explicitly with these issues in the context of nontarget effects. General methods used were either to expose the agent under outdoor conditions or to assess the agent’s lethal temperature. All methods available are summarized in Boivin et al. (Chapter 6, this volume). A quite different approach that may be useful in predicting the likelihood of establishment of a biological control agent is based on ecoregions (see Cock et al., Chapter 12, this volume). However, even when an exotic biological control agent is not able to establish permanently, seasonal persistence might be possible. This means that potential nontarget effects would be limited in time and dependent on the dispersal abilities of the agent. Despite the large amount of literature on dispersal in general, few studies have been carried out on dispersal of biological control agents specifically to assess non-target effects. The most important details required include the numbers leaving release fields (or the greenhouse), and the densities of agents at certain distances

from the point of release. Suitable methods to gather these data are provided by Mills et al. (Chapter 7, this volume).

Modelling In addition to experimental studies, modelling approaches can also be used to predict potential risks of biological control agent introductions. Using a Nicholson-Bailey model, Lynch et al. (2002) studied whether transient non-target effects can occur at an early stage of a biological control introduction due to the very high target and, consequently, agent populations. Interestingly, this study demonstrated the potential for a strong, transient decline of a non-target host population, even when the biological control agent has a very low acceptance of the non-target species. Recently, another modelling study was conducted with the aim of making predictions for populations of non-targets when these suffered from, for example, 15% parasitism (Barlow et al., 2004). Building upon the vast amount of knowledge on Microctonus spp. introduced in New Zealand, Barlow et al. (2004) used discrete Ricker or continuous logistic models that incorporated density dependence and the intrinsic rate of increase as the key factors. Using the same parasitism rate, the model predicted reductions of two nontarget host populations of 8% or 35%, respectively, and the major factor was found to be the intrinsic rate of increase of populations at different altitudes. We believe that such studies are potentially important in addressing the risks of biological control agents to non-target populations, but on the other hand we feel that the special value of modelling studies will become apparent only when these have been validated with field, or at least experimental, data.

General Considerations Regarding the Regulation of Invertebrate Biological Control Agents Above, we have provided a short overview on the status of non-target testing of arthro-


pod biological control with special emphasis on methodological aspects. We also identified several difficulties encountered and briefly discussed them where appropriate. However, some more general constraints may be important to note as well. First, the statistical analysis of studies testing for non-target effects is sometimes inappropriate (see Hoffmeister et al., Chapter 13, this volume). For instance, there is not enough discussion on the number of replicates that should be carried out in host range testing; often this number is too low. A still unsolved issue is the question of whether one or very few replicates showing negative results are sufficient to conclude that the non-target host is outside of the agent’s host range. Clearly, statistical power decreases if a small number of replicates are carried out, and low power may be especially critical in the context of risk assessment (see Hoffmeister et al., Chapter 13, this volume). Another problem, especially valid for many field studies, is that they often have been limited in time (e.g. one field season only) and space. Longer-term studies may allow more precise conclusions to be drawn on non-target impacts, but have rarely been conducted in the past. The importance of spatial dimension was demonstrated by Follett et al. (2000), who found non-target parasitism to be dependent on the elevation level of Hawaiian Islands. Clearly, to increase the temporal or spatial scale of such studies would increase the costs, a problem that is discussed below. In those cases where parasitism/predation of a non-target host was observed, it is important to know the consequences at the population level. However, impact of biological control agents on field populations of non-target species has rarely been investigated experimentally. Even if effects on the population level have been demonstrated, there is still no consensus as to what a relevant non-target effect is. First approaches have been outlined by Lynch et al. (2001), who suggested a severity index ranging from zero (no negative reports) to nine (large-scale extinction). They concluded that few serious non-target effects

9

were observed if the baseline is the extirpation of host populations on regional or even larger scales. However, biological control agents will already be rejected at a lower level of impact; but at what level of effect to reject a natural enemy is an important and yet unsolved question. As host specificity again (and establishment for inundative releases) will be the central issue(s), the question may finally be: how many non-target species should be in the host range of a biological control agent in order to consider it unsafe (see van Lenteren et al., Chapter 3, this volume)? What about an agent that has the potential to attack some non-target species, but on the other hand also has the potential for large benefits? We believe that such questions will be of increasing importance in the risk assessment of biological control agents and these questions are being addressed by Bigler and Kölliker-Ott (Chapter 16, this volume). One disadvantage of regulating invertebrate biological control agents would be the increased costs and time lag to bring new biological control agents on to the market. While producers of biological control agents must invest more initially to develop new agents, these costs are likely to be passed to growers who buy biological control agents, and ultimately to consumers who want to purchase ‘pesticide-free’ products. There is also the risk that a few producers of biological control agents will dominate the biological control industry and small units will be eliminated. On the other hand it is important that augmentative biological control is not oversold; that is, recommended when unnecessary or when not appropriate. A spin-off benefit of regulating biological control agents will be the increased difficulty of selling products that are ineffective or inappropriate, and which may nevertheless pose risks to the environment. Another potential benefit would be greater protection of intellectual property. Thus, regulations would enhance reputable biological control agent manufacturers and sellers, make biological control more science-based and help to maintain a good image of biological control by the public.

10


Given the limited resources available for biological control projects, it was stated that extensive assessment of non-target effects would be unrealistic and impractical (Messing, 2001). If a large number of species are tested with detailed investigations of the host-finding behaviour, and tests are conducted under semi-field conditions, then costs can be substantial. Obviously, the most costly species are those having a relatively wide host range, and it is worth noting that for polyphagous biological control agents, such as most trichogrammatids, a comprehensive list of non-target species may not be manageable. We believe that if polyphagous agents are to be considered at all (e.g. in inundative biological control), other approaches of risk assessment may have to be used. For instance, studies on habitat specificity or dispersal might be more promising than pure host range testing to determine the risk of such agents. One example of what can be done to assess non-target effects of the polyphagous T. brassicae was recently provided by Babendreier et al. (2003b). Another example is nematodes, which are often not restricted in their host range, but hardly any non-target effects due to the release of nematodes have been observed in the past (see Barratt et al., Chapter 10, this volume). If risks are not negligible, a cost–benefit analysis will provide a more accurate and balanced picture of the advantages and disadvantages of releasing an agent; in fact, information on potential benefits is also required by the OECD guidance document (Table 1.1), but to date only limited information on cost–benefit analysis in invertebrate biological control is available. One of the few papers including such information in the context of biological control was recently published by Heimpel et al. (2004) on the risks and benefits of introducing parasitoids for control of soybean aphids. In this book, we shall try to elaborate further on this issue (see Bigler and Kölliker-Ott, Chapter 16, this volume). Regulations will certainly have an impact on the business strategy of biological control manufacturers, particularly when generalist species are involved. Investigations of local

strains of the same or a related species could be encouraged. However, local populations should be used only as source material for laboratory cultures, not as a convenient supply. In North America, the convergent ladybird beetle, Hippodamia convergens Guérin-Meneville, is collected from overwintering aggregations and shipped directly to buyers (Gillespie et al., 2002). This is questionable because this practice has the potential for reducing local biodiversity and for transmitting contaminants (e.g. parasitoids and diseases) to native species in the area of release (see Goettel and Inglis, Chapter 9, this volume). A problem somehow specific to the OECD guidance document is the fact that it often requests ‘available information’. There will immediately be the question of what to do if there are no data available for a specific question. Moreover, the OECD document includes some issues that have received little attention in the past, including the potential for interbreeding (see Hopper et al., Chapter 5, this volume), the potential of damage to non-target plants (see Albajes et al., Chapter 8, this volume) or the potential risk that a biological control agent carries unwanted contaminants (see Goettel and Inglis, Chapter 9, this volume). These topics are addressed in the book and information on how to tackle such questions is provided.

Conclusions Despite the fact that few non-target effects associated with arthropod biological control have been reported, the number of studies that have tested for such effects increased substantially during the last decade. A lot of progress has been achieved and many recent introductions have been accompanied by appropriate host range assessments. Nevertheless, we are still not at the stage where host-range assessment combined with pre- and post-release studies are standard procedures in each biological control project, a suggestion put forward by Barratt et al. (Chapter 10, this volume). We would also like to stress that often only a fraction


of all potential risks have been assessed. This becomes especially obvious when looking at indirect effects where it is clearly not possible to test for all interactions. Although the above-mentioned efforts have already led to increased costs of biological control projects, a recent evaluation of the IPPC Code of Conduct revealed no decrease in the number of introductions of exotic biological control agents, but rather indicated a delay of introductions (Kairo et al., 2003). This, however, may be due to the fact that the IPPC Code was not legally binding to involved parties. If guidance documents (e.g. the OECD document) could find their way into national laws, then this situation may change in the future (i.e. the number of biological control projects and introductions might decrease). However, the application of appropriate regulatory procedures is important in order to maintain public confidence in biological control and to facilitate introductions and the commercial use of exotic biological control agents in the future.

11

Since regulation and non-target testing will increase associated costs, it is important to use the available resources as efficiently as possible. Therefore, it is important to provide guidance on testing non-target effects, and for this goal it is extremely valuable to have appropriate methods available. This book aims to contribute to both objectives. We believe that improving non-target testing procedures in arthropod biological control is not only necessary for reducing the potential of adverse effects on non-targets even further, but also for preventing the hurdles that accompany over-regulation. The vast majority of agents used in arthropod biological control have been shown to be safe. Finally, we suggest conducting careful and well-balanced analyses of potential risks and benefits for biological control projects in the future, keeping in mind that all plant protection methods bear risks and benefits which need to be evaluated against each other.

References Babendreier, D., Rostas, M., Hofte, M.C.J., Kuske, S. and Bigler, F. (2003a) Effects of mass releases of Trichogramma brassicae on predatory insects in maize. Entomologia Experimentalis et Applicata 108, 115–124. Babendreier, D., Schoch, D., Kuske, S., Dorn, S. and Bigler, F. (2003b) Non-target habitat exploitation by Trichogramma brassicae (Hym.: Trichogrammatidae): what are the risks for endemic butterflies? Agricultural and Forest Entomology 5, 199–208. Babendreier, D., Bigler, F. and Kuhlmann, U. (2005) Methods used to assess non-target effects of invertebrate biological control agents of insect pests. BioControl 50, 821–870. Barlow, N.D., Barratt, B.I.P., Ferguson, C.M. and Barron, M.C. (2004) Using models to estimate parasitoid impacts on non-target host abundance. Environmental Entomology 33, 941–948. Barratt, B.I.P. (2004) Microctonus parasitoids and New Zealand weevils: comparing laboratory estimates of host ranges to realized host ranges. In: Van Driesche, R.G and Reardon, R. (eds) Assessing Host Ranges for Parasitoids and Predators Used for Classical Biological Control: a Guide to Best Practice. FHTET-2004-03, Forest Health Technology Enterprise Team, Morgantown,West Virginia, USA, pp. 103–120. Barratt, B.I.P., Evans, A.A., Ferguson, C.M., Barker, G.M., McNeill, M.R. and Phillips, C.B. (1997) Laboratory non-target host range of the introduced parasitoids Microctonus aethiopoides and M. hyperodae (Hymenoptera: Braconidae) compared with field parasitism in New Zealand. Environmental Entomology 26, 694–702. Barron, M.C., Barlow, N.D. and Wratten, S.D. (2003) Non-target parasitism of the endemic New Zealand Red Admiral Butterfly (Bassaris gonerilla) by the introduced biological control agent Pteromalus puparum. Biological Control 27, 329–335. Bigler, F., Bale, J., Cock, M., Dreyer, H., GreatRex, R., Kuhlmann, U., Loomans, A. and van Lenteren, J. (2005) Guideline on information requirements for import and release of invertebrate biological control agents in European countries. Biocontrol News and Information 26, 115N-123N. Brown, M.W. (2003) Intraguild responses of aphid predators on apple to the invasion of an exotic species, Harmonia axyridis. BioControl 48, 141–153.

12


Burgio, G., Santi, F. and Maini, S. (2002) On intra-guild predation and cannibalism in Harmonia axyridis (Pallas) and Adalia bipunctata L. (Coleoptera: Coccinellidae). Biological Control 24, 110–116. Coombs, M. (2003) Post-release evaluation of Trichopoda giacomellii (Diptera: Tachinidae) for efficacy and non-target effects. In: Van Driesche, R.G. (ed.) Proceedings of the 1st International Symposium on Biological Control of Arthropods, Honolulu, Hawaii, 14–18 January 2002. FHTET-2003-05, United States Department of Agriculture, Forest Service, Morgantown, West Virginia, USA, pp. 399–406. De Nardo, E.A.B. and Hopper, K.R. (2004) Using the literature to evaluate parasitoid host ranges: a case study of Macrocentrus grandii (Hymenoptera: Braconidae) introduced into North America to control Ostrinia nubilalis (Lepidoptera: Crambidae). Biological Control 31, 280–295. Denoth, M., Frid, L. and Myers, J.H. (2002) Multiple agents in biological control: improving the odds? Biological Control 24, 20–30. Duan, J.J. and Messing, R.H. (2000) Evaluating non-target effects of classical biological control: fruit fly parasitoids in Hawaii as a case study. In: Follett, P.A. and Duan, J.J. (eds) Nontarget Effects of Biological Control. Kluwer Academic Publishers, Norwell, Massachusetts, USA, pp. 95–109. EPPO (1997) EPPO/CABI workshop on safety and efficacy of biological control agents in Europe. EPPO Bulletin 27, 1–3. EPPO (1999) First import of exotic biological control agents for research under contained conditions. EPPO Bulletin 29, 271–272. EPPO (2001) Import and release of exotic biological control agents. EPPO Bulletin 31, 33–35. EPPO (2002) List of biological control agents widely used in the EPPO region. EPPO Bulletin 32, 447–461. Follett, P.A., Duan, J.J., Messing, R.H. and Jones, V.P. (2000) Parasitoid drift after biological control introductions: Re-examining Pandora’s Box. American Entomologist 46, 82–94. Gariepy, T.D., Kuhlmann, U., Haye, T., Gillott, C. and Erlandson, M. (2005) A single-step multiplex PCR assay for the detection of European Peristenus spp. (Hymenoptera: Braconidae), parasitoids of Lygus spp. (Hemiptera: Miridae). Biocontrol Science and Technology, 15, 481–495. Gillespie, D.R., Shipp, J.L., Raworth, D.A. and Foottit, R.G. (2002) Aphis gossypii Glover, melon/ cotton aphid, Aulacorthum solani (Kaltenbach), foxglove aphid, Macrosiphum euphorbiae (Thomas), potato aphid, and Myzus persicae (Sulzer), green peach aphid (Homoptera: Aphididae). In: Mason, P.G. and Huber, J.T. (eds) Biological Control Programmes in Canada 1981–2000. CABI Publishing, Wallingford, UK, pp. 44–49. Haye, T. (2004) Studies on the ecology of European Peristenus spp. (Hymenoptera: Braconidae) and their potential for the biological control of Lygus spp. (Hemiptera: Miridae) in Canada. PhD thesis, University of Kiel, Germany. Haye, T., Goulet, H., Mason, P.G. and Kuhlmann, U. (2005) Does fundamental host range match ecological host range of Lygus plant bug parasitoids? A retrospective case study. Biological Control, 35, 55–67. Heimpel, G.E., Ragsdale, D.W., Venette, R., Hopper, K.R., O’Neil, R.J., Rutledge, C.E. and Wu, Z.S. (2004) Prospects for importation biological control of the Soybean aphid: anticipating potential costs and benefits. Annals of the Entomological Society of America 97, 249–258. Hoddle, M.S. (2004) Analysis of fauna in the receiving area for the purpose of identifying native species that exotic natural enemies may potentially attack. In: Van Driesche, R.G. and Reardon, R. (eds) Assessing Host Ranges for Parasitoids and Predators Used for Classical Biological Control: a Guide to Best Practice. Forest Health Technology Enterprise Team, Morgantown, West Virginia, USA, pp. 24–39. Howarth, F.G. (1983) Classical biological control: panacea or Pandora’s box. Proceedings of the Hawaiian Entomological Society 24, 239–244. Howarth, F.G. (1991) Environmental impacts of classical biological control. Annual Review of Entomology 36, 485–509. IPPC (1996) Code of conduct for the import and release of exotic biological control agents. Publication No. 3, FAO, Rome, Italy. IPPC (International Plant Protection Convention) (2005) Guidelines for the export, shipment, import and release of biological control agents and other beneficial organisms. International Standards for Phytosanitary Measures No. 3. https://www.ippc.int/servlet/CDSServlet?status=ND0xMz M5OS43NjA0NyY2PWVuJjMzPXB1YmxpY2F0aW9ucyZzaG93Q2hpbGRyZW49dHJ1ZSYzNz1p bmZv#koinfo (accessed 16 November 2005).


13

Johnson, M.T., Follett, P.A., Taylor, A.D. and Jones, V.P. (2005) Impacts of biological control and invasive species on a non-target native Hawaiian insect. Oecologia 142, 529–540. Kairo, M.T.K., Cock, M.J.W. and Quinlan, M.M. (2003) An assessment of the use of the Code of Conduct for the Import and Release of Exotic Biological Control Agents (ISPM No. 3) since its endorsement as an international standard. Biocontrol News and Information 24, 15N–27N. Keller, M. (1999) Understanding host selection behaviour: the key to more effective host specificity testing. In: Withers, T.M. and Stanley, J.N. (eds) Host Specificity Testing in Australasia: Towards Improved Assays for Biological Control. CRC for Tropical Pest Management, Brisbane, Australia, pp. 84–92. Kuhlmann, U., Mason, P.G. and Foottit, R.G. (2000) Host specificity assessment of European Peristenus parasitoids for classical biological control of native Lygus species in North America: use of field host surveys to predict natural enemy habitat and host ranges. In: Van Driesche, R.G., Heard, T.A., McClay, A.S. and Reardon, R. (eds) Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: the Biological Basis for Improvement in Safety. Xth International symposium on Biological Control of Weeds, July 4–14, 1999, Bozeman, Montana. Bulletin, FHTET-99–1, USDA Forest Service Morgantown, West Virginia, USA, pp. 84–95. Lockwood, J.A., Howarth, F.G. and Purcell, M.F. (2001) Balancing Nature: Assessing the Impact of Importing Non-Native Biological Control Agents (an International Perspective). Thomas Say Publications in Entomology, ESA. Lanham, Maryland, USA, 130 pp. Louda, S.M., Pemberton, R.W., Johnson, M.T. and Follett, P.A. (2003) Nontarget effects – the Achilles’ Heel of biological control? Retrospective analyses to reduce risk associated with biocontrol introductions. Annual Review of Entomology 48, 365–396. Lynch, L.D., Hokkanen, H.M.T., Babendreier, D., Bigler, F., Burgio, G., Gao, Z.H., Kuske, S., Loomans, A., Menzler-Hokkanen, I., Thomas, M.B., Tommasini, G., Waage, J.K., van Lenteren, J.C. and Zeng, Q.-Q. (2001) Insect biological control and non-target effects: a European perspective. In: Wajnberg, E., Scott, J.K. and Quimby, P.C. (eds) Evaluating Indirect Ecological Effects of Biological Control. CABI Publishing, New York, USA, pp. 99–125. Lynch, L.D., Ives, A.R., Waage, J.K., Hochberg, M.E. and Thomas, M.B. (2002) The risks of biocontrol: transient impacts and minimum non-target densities. Ecological Applications 12, 1872–1882. Mansfield, S. and Mills, N.J. (2004) A comparison of methodologies for the assessment of host preference of the gregarious egg parasitoid Trichogramma platneri. Biological Control 29, 332–340. Messing, R.H. (2001) Centrifugal phylogeny as a basis for non-target host testing in biological control: Is it relevant for parasitoids? Phytoparasitica 29, 187–190. Murdoch, W.W., Briggs, C.J. and Nisbet, R.M. (1996) Competitive displacement and biological control in parasitoids: a model. American Naturalist 148, 807–826. NAPPO (2000) Guidelines for petition for release of exotic entomophagous agents for the biological control of pests. Secretariat of North American Plant Protection Organization, Ottawa, Canada. Neuenschwander, P. (2001) Biological control of the cassava mealybug in Africa: A review. Biological Control 21, 214–229. OECD (2003) Guidance for information requirements for regulations of invertebrates as biological control agents. OECD Environment, Health and Safety Publications. Series on Pesticides 21, 22 pp. Reitz, S.R. and Trumble, J.T. (2002) Competitive Displacement among insects and arachnids. Annual Review of Entomology 47, 435–465. Sands, D. (1997) The ‘safety’ of biological control agents: Assessing their impact on beneficial and other non-target hosts. Memoirs of the Museum of Victoria 56, 611–616. Sands, D. (1998) Guidelines for testing host specificity of agents for biological control of arthropod pests. In: Zalucki, M.P., Drew, R.A.I. and White, G.G. (eds). Proceedings of the Sixth Australasian Applied Entomological Research Conference, Volume 1. University of Queensland Press, Brisbane, Australia, pp. 556–560. Sands, D.P.A. and Van Driesche, R.G. (2000) Evaluating the host range of agents for biological control of arthropods: rationale, methodology and interpretation. In: Van Driesche, R.G., Heard, T.A., McClay, A.S. and Reardon, R. (eds) Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: the Biological Basis for Improvement in Safety. Xth International symposium on Biological Control of Weeds, July 4–14, 1999, Bozeman, Montana. Bulletin, FHTET-99-1, USDA Forest Service Morgantown, West Virginia, USA, pp. 69–83.

14


Sands, D.P.A. and Van Driesche, R.G. (2004) Using the scientific literature to estimate the host range of a biological control agent. In: Van Driesche, R.G. and Reardon, R. (eds) Assessing Host Ranges for Parasitoids and Predators Used for Classical Biological Control: a Guide to Best Practice. Forest Health Technology Enterprise Team, Morgantown, West Virginia, USA, pp. 15–23. Schellhorn, N.A., Kuhman, T.R., Olson, A.C. and Ives, A.R. (2002) Competition between native and introduced parasitoids of aphids: non-target effects and biological control. Ecology 83, 2745–2757. Simberloff, D. and Stiling, P. (1996) How risky is biological control? Ecology 77, 1965–1974. Symondson, W.O.C. (2002) Molecular identification of prey in predator diets. Molecular Ecology 11, 627–641. Van Driesche, R.G. and Hoddle, M. (1997) Should arthropod parasitoids and predators be subject to host range testing when used as biological control agents? Agriculture and Human Values 14, 211–226. Van Driesche, R.G. and Murray, T.J. (2004a) Overview of testing schemes and designs used to estimate host ranges. In: Van Driesche, R.G. and Reardon, R. (eds) Assessing Host Ranges for Parasitoids and Predators Used for Classical Biological Control: a Guide to Best Practice. Forest Health Technology Enterprise Team, Morgantown, West Virginia, USA, pp. 68–89. Van Driesche, R.G. and Murray, T. J. (2004b) Parameters used in laboratory host range tests. In: Van Driesche, R.G. and Reardon, R. (eds) Assessing Host Ranges for Parasitoids and Predators Used for Classical Biological Control: a Guide to Best Practice. Forest Health Technology Enterprise Team, Morgantown, West Virginia, USA, pp. 56–67. Van Driesche, R.G. and Reardon, R. (2004) Assessing Host ranges for Parasitoids and Predators Used for Classical Biological Control: a Guide to Best Practice. Forest Health Technology Enterprise Team, Morgantown, West Virginia, USA. van Lenteren, J.C., Babendreier, D., Bigler, F., Burgio, G., Hokkanen, H.M.T., Kuske, S., Loomans, A.J.M., Menzler-Hokkanen, I., van Rijn, P.C.J., Thomas, M.B., Tommasini, M.G. and Zeng, Q.-Q. (2003) Environmental risk assessment of exotic natural enemies used in inundative biological control. BioControl 48, 3–38. Wang, X.G. and Messing, R.H. (2002) Newly imported larval parasitoids pose minimal competitive risk to extant egg-larval parasitoid of tephritid fruit flies in Hawaii. Bulletin of Entomological Research 92, 423–429. Whithers, T.M. and Browne, L.B. (2004) Behavioral and physiological processes affecting outcomes of host range testing. In: Van Driesche, R.G. and Reardon, R. (eds) Assessing Host ranges for Parasitoids and Predators Used for Classical Biological Control: a Guide to Best Practice. Forest Health Technology Enterprise Team, Morgantown, West Virginia, USA, pp. 40–55. Zilahi-Balogh, G.M.G., Kok, L. and Salom, S. (2002) Host specificity of Laricobius nigrinus Fender (Coleoptera: Derodontidae), a potential biological control agent of the hemlock woolly adelgid, Adelges tsugae Annand (Homoptera: Adelgidae). Biological Control 24, 192–198.

2

Selection of Non-target Species for Host Specificity Testing

Ulrich Kuhlmann,1 Urs Schaffner1 and Peter G. Mason2 1CABI

Bioscience Switzerland Centre, Rue des Grillons 1, 2800 Delémont, Switzerland (email: [email protected]; fax number: +41-32-4214871); 2Agriculture and Agri-Food Canada, Research Centre, Central Experimental Farm, Ottawa, Ontario, K1A 0C6 Canada (email: [email protected]; fax number: +1-613-7591701)

Abstract We present comprehensive recommendations for setting up test species lists for arthropod biological control programmes that are scientifically based and ensure that all aspects of potential direct impacts are considered. It is proposed that a set of categories, including ecological similarities, phylogenetic/taxonomic affinities and safeguard considerations are applied to ecological host range information to develop an initial test list. This list is then filtered to reduce the number of species to be tested by eliminating those with different spatial, temporal and morphological attributes and those species that are not readily obtained, and thus unlikely to yield scientifically sound data. The revised test list is used for the actual testing but can (and should) be revised if new information obtained indicates that additional or more appropriate species should be included. Use of the recommendations is illustrated by a case study on the host specificity of a tachinid fly Celatoria compressa Wulp, a candidate for use as a biological control agent against the western corn rootworm, Diabrotica virgifera virgifera LeConte.

Introduction Biological control is an environmentally friendly and highly cost-effective strategy for combating pests in agriculture and forest ecosystems. Despite recent concerns about unintended effects, the use of exotic natural enemies against invasive alien species in natural and agricultural habitats remains a key component of integrated pest management. What has changed during the last decade is the importance of scientifically sound decisions for ensuring that exotic biological control agents introduced

into new environments have minimal impact on non-target species. Host-specificity testing of entomophagous biological control agents has lagged behind that of phytophagous biological control agents. In fact, until the warnings by Howarth (1983, 1991), Lockwood (1993a,b, 2000) and Louda et al. (1997), concerns about impacts on non-target species were infrequently considered in entomophagous biological control projects. Lynch et al. (2001) reviewed the published and unpublished European information and determined that a mere 1.5% of entomophagous biological


15

16

U. Kuhlmann et al.

control agents introduced before 1999 appeared to have undergone host specificity analyses, thus the extent of information on non-target impacts, including selection of species to be tested, is limited. Selection of appropriate species for testing potential impacts of candidate biological control agents is the first critical step in the process once the need for pest suppression is justified and one or more potential agents have been identified. Several authors, e.g. Sands (1997, 1998); van Lenteren et al. (2003), have suggested that the centrifugal phylogenetic method of Wapshere (1974) should be the primary method used for selecting non-target species for testing candidate entomophagous biological control agents. However, the centrifugal phylogenetic approach may not always be feasible because of taxonomic uncertainties and the greater number of taxa that could be required in testing compared to weeds (Kuhlmann et al., 2000). Moreover, other parameters such as the feeding niche or the common habitat of target and non-target species may be more meaningful, at least for certain biological control agents (Messing, 2001). In this chapter, we review the current practice of developing test plant lists in weed biological control programmes as a basis for discussion, what determines parasitoid host ranges, and review the approaches taken in recent arthropod biological control programmes. We propose comprehensive recommendations for setting up test species lists for arthropod biological control programmes that are scientifically based and ensure that all aspects of potential direct impacts are considered. At the same time, the recommendations attempt to take into consideration possible practical constraints associated with arthropod host specificity screening (Sands and Van Driesche, 2000). Use of the recommendations is illustrated by a case study on the host specificity of a tachinid fly, Celatoria compressa, a candidate for use as a biological control agent against the chrysomelid Diabrotica virgifera virgifera (Kuhlmann et al., 2005). Although most of

the available information is on parasitoids, the recommendations developed should apply to other invertebrate groups such as arthropod predators and entomopathogenic nematodes.

What can be Learned from Current Practice in Weed Biological Control? For more than 30 years, the screening of the fundamental (= physiological) and the ecological host range of candidate biological control agents has been the most crucial step in pre-release studies of any weed biological control programme (Harris and Zwoelfer, 1968; Zwoelfer and Harris, 1971; Wapshere, 1974). Because of the overriding importance of safety, greatest care is taken in selecting appropriate test plants and in designing meaningful screening tests to accurately predict the host specificity of potential control agents. Host range studies were originally developed to protect agricultural crops from unwanted attack. While taxonomic relatedness provides a starting point, in practice other considerations, such as inclusion of beneficial (i.e. crop) species and those that are aesthetically important (i.e. species at risk), are also considered. At present, the selection of test plants is based on proposals made by Harris and Zwoelfer (1968) Wapshere (1974) and Wapshere (1989). The aim is to select those plant species most likely to be hosts of the organism in question, without undue expansion of the test plant list. The basis of the standard selection protocol is the centrifugal phylogenetic method developed by Wapshere (1974). This method was based on the observation that the host range of specialist herbivores is usually restricted to one or a few phylogenetically related plant taxa. Recent studies have confirmed this pattern for many, but not all, insect herbivore groups (Bernays, 2000; Pemberton, 2000; for groups including biological control agents see Dobler, 2001; Ronquist and Liljeblad, 2001). The centrifugal phylogenetic method involves selecting and testing plants of increasingly distant phylogenetic


relationship to the target weed (Wapshere, 1974; Table 2.1). As a safeguard against failure of the centrifugal phylogenetic method, Wapshere (1974) proposed adding a number of economically important plants to the test plant list, as well as any plant species on which the candidate agent had previously been recorded. In modern weed biological control programmes, additional plants considered in test plant lists are species with phytochemical or morphological features similar to those of the target host; plants known to be attacked by organisms closely related to the candidate biological agent; threatened and endangered species in the same family as the target species; and those occurring in the same habitat (Table 2.2).

17

As noted by Briese et al. (2002), while the centrifugal phylogenetic method claims to be phylogenetically based, it used to be – and largely still is to date – based on taxonomic circumscription. For example, it is only recently that comprehensive phylogenies of the species-rich genera Centaurea and Senecio, both of which include invasive weeds, and species in the closely related genera, have been hypothesized using molecular data (Garcia-Jacas et al., 2001; Pelser et al., 2002). The number of plant species that should be included in a non-target list depends mainly on: ● The taxonomic position of the target weed – whether it belongs to an isolated family or to a family with close relations.

Table 2.1. Wapshere’s (1974) centrifugal phylogenetic testing method. Testing sequence

Plants to be tested

Host range determined if plants at that phylogenetic level remain unattacked

1st 2nd 3rd 4th 5th 6th

Other forms (ecotypes/biotypes) of target species Other species of same genus Other members of tribe Other members of subfamily Other members of family Other members of order

Specific to clone Specific to species Specific to genus Specific to tribe Specific to subfamily Specific to family

Table 2.2. Plant categories listed in the Reviewer’s Manual for the Technical Advisory Group for Biological Control of Weeds (USDA/APHIS, Plant Protection and Quarantine) for compilation of a test plant list. Category 1: Category 2: Category 3:

Category 4: Category 5:

Category 6:

Category 7:

Genetic types of the target pest species (genotypes, geographic populations, etc.) Species in the same genus as the target weed, divided by subgenera (if applicable), including economically and environmentally important plants of North America. Species in other genera in the same family as the target weed, divided by subfamily (if applicable), including economically and environmentally important plants of North America. Threatened and endangered species in the same family as the target weed, divided by subgenus, genus and subfamily. Species in other families in the same order that have some phylogenetic, morphological or biochemical similarities to the target weed, or that share the same habitat, including economically and environmentally important plants of North America. Species in other orders that have some morphological or biochemical similarities to the target or that share the same habitat, including economically and environmentally important plants of North America. Any plant species on which the biological control agent or its close relatives (within the same genus) have previously been found or recorded feeding and/or reproducing.

18

U. Kuhlmann et al.

● The number of closely related cultivated plants, and other-valued wild plants. ● The geographic and/or ecological isolation of the release area. ● Whether or not the candidate biological control organism belongs to a systematic group which is known to be restricted to a small group of closely related plants (genus, subtribe and tribe). In recent examples of host range determination, the number of plant species screened ranges from 40 to more than 100. In host plant lists of modern weed biological control projects, unrelated plant species sharing conspicuous secondary metabolites or morphological characters are represented to a certain extent, but it is usually not known whether the characters selected are indeed of relevance in the host selection behaviour of the candidate weed biological control agent. To increase the chances of detecting disjunct oligophagy, one needs to elucidate the cues used by the candidate species in selecting and accepting host plants (Schaffner, 2001). In a recent review of the relevance of the criteria set up by Wapshere (1974), Briese et al. (2002) argued that none of the safeguard criteria has generated additional insight into the results obtained by applying the centrifugal phylogenetic method. Briese et al. (2002) therefore recommended dropping these safeguard criteria to reduce costs of pre-release studies. However, host range testing with the agromyzid fly Napomyza sp. near lateralis in a biological control project against Russian knapweed, Acroptilon repens (L.) de Candolle, revealed that the only plant species outside the knapweeds (genera Acroptilon and Centaurea) found to be both within the fundamental and ecological host ranges of this species is the distantly related host plant of a sibling agromyzid species (U. Schaffner, Delémont, 2004, unpublished results). Since the addition of a few safeguard species, e.g. in nochoice feeding bioassays, usually does not cause major additional costs in weed biological control programmes, further inclusion of safeguard species may be scientifically and politically justified, despite the fact that they rarely contribute to additional informa-

tion on the host affiliation. While these practices have been developed over time for weed biological control, in arthropod biological control other factors may determine the parasitoid host range. These factors will be outlined in the following section.

What Determines Parasitoid Host Range? Our knowledge of parasitoid host ranges is based primarily on associations made through rearing a limited number of host species. The number of studies exploring the evolutionary and ecological determinants of host use in parasitoids is growing (e.g. Hawkins, 1994; Hawkins and Sheehan, 1994), yet for most groups we have limited information on the relative importance of host habitat, processes of host location, physiological interactions with hosts, host defences or host phylogenetic history in influencing parasitoid host ranges (Stireman and Singer, 2003). Documenting parasitoid host range is far more difficult than collecting data on host parasitoid species load because it involves rearing parasitoids to the adult stage for identification from many different species of hosts rather than rearing or dissecting many individuals of a single host species. Existing data are of two main types: large catalogues of known host associations, frequently, although not always, concerned with selected taxonomic groups of parasitoids and very seldom including any quantitative information; and food webs containing information on all parasitoids attacking a restricted range of hosts, often in a single geographical area (Memmott and Godfray, 1993). Information from host catalogues must be treated with extreme caution (Askew and Shaw, 1986; Noyes, 1994). The difficulties of parasitoid taxonomy, plus the risk of erroneous parasitoidhost associations, render many large catalogues almost useless for ecological studies, but exceptions occur where experts have at least carefully scrutinized host records in the literature (e.g. Boucek and Askew, 1968; Griffiths, 1964–1968).


No parasitoid successfully parasitizes all hosts in the environment, and species that are attacked by the same parasitoid share certain characteristics. The two most important determinants of host range are most probably host taxonomy and shared ecology (Askew and Shaw, 1986; Shaw, 1988). The correlation between host taxonomy and parasitoid range has been demonstrated on numerous occasions (e.g. Askew, 1961; Griffiths, 1964–1968) and these correlations can arise for at least two reasons. First, parasitoids may attack closely related hosts because they share similar physiological properties and defence mechanisms. Second, closely related parasitoids are likely to be biologically similar, for example, they are more likely to feed on hosts using the same host plant or to have similar feeding niches. The importance of shared ecology is best illustrated by examples of unrelated hosts of parasitoids that share host plants or feeding niches and are attacked by the same parasitoid. Hosts that feed on the same food plant frequently share the same parasitoids (e.g. Vinson, 1981, 1985; Fitton et al., 1988). Plant chemistry may influence parasitoid host range if hosts sequester toxins from their food plants. Chemical similarity is known to influence polyphagy at the herbivore trophic level, and chemical diversity has been linked with host range (e.g. Strong et al., 1984), including semiochemicals released when the plant is damaged by herbivore hosts (e.g. Godfray, 1994). Hoffmeister (1992) surveyed the parasitoids attacking seven races or species of tephritid fly feeding in the fleshy seeds of a variety of trees, shrubs and climbers in Europe. He found that host ecology, broadly defined as phenology, feeding habitat and host plant taxonomy, was more important than host taxonomy in determining the make-up of the parasitoid complex. Based on Godfray (1994), some predictions can be made about the relative host ranges of parasitoid species with an intimate biochemical and physiological connection with their hosts (larval koinobiont endoparasitoids), species that do not have to contend with active host defences (larval

19

idiobiont ectoparasitoids) and species that attack non-growing host stages. The first group should be relatively specialized and their host range will be strongly influenced by host taxonomy. The last group should be less specialized and their host range will be influenced by both host taxonomy and host ecology. Thus: ● Koinobionts should have fewer hosts than idiobionts. ● Pupal and egg and adult parasitoids should be less specialized than larval parasitoids (Strand, 1986). ● The koinobiont parasitoids of taxonomically isolated hosts should attack few other species. ● The idiobiont parasitoids of ecologically isolated hosts should attack few other species. Idiobiont larval parasitoids more often attack hosts in concealed feeding niches where death or permanent paralysis is less likely to increase the risk of predation (Hawkins, 1990). There will be numerous exceptions to the broad generalizations set up by Godfray (1994). For example, many tachinid flies are koinobiont endoparasitoids, yet can subvert the host immune system of a wide variety of species and thus enjoy a remarkably broad host range (Belshaw, 1994). The suggestion that koinobionts have broader host ranges than idiobionts has some empirical support. In surveys of parasitoids of lepidopteran and hymenopteran leafminers, Askew and Shaw (1986), Pschorn-Walcher and Altenhofer (1989) and Sato (1990) all observed more restricted host ranges among idiobionts than among koinobionts. The importance of shared ecology should not be overemphasized. There are many examples of parasitoids that attack one or a few closely related hosts in a wide variety of habitats (e.g. Price, 1981). Futuyma and Moreno (1988) reviewed a variety of macroevolutionary aspects of parasitoid host range. In some taxa, particular specializations appear to be taxonomically conserved: all Eucharitidae parasitize ants; the complete Opiinae and Alysiinae clade (Braconidae) are restricted to cyclorraphous

20

U. Kuhlmann et al.

Diptera; and the ichneumonid subfamily Ichneumoninae and the braconid subfamily Microgasterinae parasitize only Lepidoptera (Futuyma and Moreno, 1988). In other groups, for example the Eulophidae and Pteromalidae, nearly all species parasitize a restricted set of hosts, yet the clade is not committed to any particular host group. Generalism also may be phylogenetically conserved. The braconid genus Dacnusa is comprised of many species specialized on particular agromyzid leaf miners, but the few species with wide host ranges are closely related (Godfray, 1994). In summary, although determining parasitoid host ranges is plagued with difficulties (Shaw, 1994), it appears that most parasitoids attack a narrow range of hosts (Memmott et al., 2000). The two principal factors that limit host ranges in parasitoids are thought to be taxonomic relatedness of hosts and host ecology. The effect of host taxonomic affinity is believed to be related primarily to physiological (and morphological) defences of hosts that may require specific adaptations of their parasitoids (Vinson and Iwantsch, 1980; Godfray, 1994). The proposal that physiological defences limit parasitoid host ranges is analogous to arguments concerning the importance of secondary chemicals in the specialization of phytophagous insects on food plants (Ehrlich and Raven, 1964). Ecological characteristics that influence host use by parasitoids include the plants on which a host feeds (Vinson, 1981; Askew, 1994), the microhabitat in which it feeds (Weseloh, 1993), the host’s phenology (Askew, 1961) and the host and parasitoid mobility (Barratt, 2004). Thus, parasitoid host range is determined by biological and ecological factors, often, but not always, associated with related host species.

What Methods Have Been Used So Far for Selecting Non-target Species in Arthropod Biological Control? A review of some recent studies suggests that in practice, criteria for selecting nontarget species for testing can be divided

into five categories: ecological, phylogenetic, socio-economic, biological and availability of test species (Table 2.3). Many studies state the reasons behind selection of the test species, and all but three studies used at least two of the categories in their selection. The numbers of non-target species tested in the laboratory ranged from one to 23 (average 10.5). Although Rutledge and Wiedenmann (1999) and Bourchier (2003) did not actually do any testing, both provided important ideas for selecting test species. Phylogenetic considerations were based on taxonomic relatedness (e.g. same genus, same family, etc.) of test species to target host. Ecological features included overlaps of geographic range, habitat preference and feeding niche of species representing different components of the community. Biological characteristics included known host range, phenological overlap of the target and nontarget species, dispersal capability of the candidate biological control agent (and parasitized host), morphological similarity, behavioural factors (e.g. feeding, oviposition, host location, etc.) and overlap of the physiological host range of biological control agents. Socio-economic factors included whether a potential test species was commercially important (e.g. a pollinator), beneficial (e.g. predator, weed biological control agent) or of conservation importance (e.g. rare or endangered). The availability of non-target material was considered, and sources included commercial or laboratory cultures, field collections and progeny of field-collected individuals. In most examples literature records provided important guidance on at least broad groups, habitats or biological parameters. In one case, surveys by Fuester et al. (2001) in the area of origin of the target species provided information on actual host range that was useful for selecting test lists. Sands et al. (1993) studied the host range of Cotesia erionotae (Wilkinson) (Hymenoptera: Braconidae), a parasitoid of the banana skipper Erionota thrax (L.) (Lepidoptera: Hesperiidae). One non-target species in the same family as the banana skipper and three species that were consid-

Table 2.3. Review of some recent studies suggests that, in practice, criteria for selecting non-target species for testing invertebrate biological control agents for arthropod pests can be divided into five categories: ecological, phylogenetic, socio-economic, biological and availability of test species. Ecological similarity

Phylogenetic affinity Sands et al. (1993)

Socio-economic

Biological

Availability

Andow et al. (1995)

–

Sands et al. (1993)

Neale et al. (1995)

Neale et al. (1995)

Neale et al. (1995)

Duan and Messing (1996, 1997)





Duan et al. (1997)

Duan et al. (1997)

Duan et al. (1997)

Duan et al. (1997)

Duan et al. (1997)

Cameron and Walker (1997) Barratt et al. (1997, 1998, 2000)

Barratt et al. (1997, 1998, 2000)

Barratt et al. (1997, 1998, 2000)

Kitt and Keller (1998)

Cameron and Walker (1997)

Cameron and Walker (1997)

Barratt et al. (1997, 1998, 2000)

Barratt et al. (1997, 1998, 2000)



Orr et al. (2000)

Orr et al. (2000)

Rutledge and Wiedenmann (1999) Porter (2000)

Porter (2000)

Boettner et al. (2000)

Porter (2000) Boettner et al. (2000)

Boettner et al. (2000)

Fuester et al. (2001)

Boettner et al. (2000) Fuester et al. (2001) Mansfield and Mills (2002)

Munro and Henderson (2002)

Munro and Henderson (2002)

Bourchier (2003)

Bourchier (2003)

Babendreier et al. (2003a,b)

Babendreier et al. (2003a,b) Babendreier et al. (2003c)


Andow et al. (1995)

Bourchier (2003) Babendreier et al. (2003a,b) Babendreier et al. (2003c)

Babendreier et al. (2003d)

Babendreier et al. (2003d) Benson et al. (2003) 21

22

U. Kuhlmann et al.

ered of commercial value were selected. Although not stated, it appears that test individuals were obtained commercially or field-collected. The results indicated that none of the non-target species would be attacked. Andow et al. (1995) developed a hypothetical analysis of risks to non-target Lepidoptera after release of Trichogramma nubilale Ertle and Davis (Hymenoptera: Trichogrammatidae) for control of Ostrinia nubilalis Hübner (Lepidoptera: Crambidae). Selection of the non-target Karner Blue Butterfly, Lycaeides melissa samuelis Nabakov, as a test species was based on endangered status, spatial occurrence, known host range of the agent, phenological overlap of the target and non-target species, dispersal of the biological control agent and mortality of the agent during dispersal. Their analysis indicated that populations of L. m. samuelis were unlikely to be reduced by inundative introductions of T. nubilale. Neale et al. (1995) developed a nontarget test list for assessing one encyrtid and two eulophid larval parasitoids of the citrus leafminer, Phyllocnistris citrella Stainton (Lepidoptera: Gracillaridae) in Australia. Although not stated, the test species were probably chosen based on ecological, phylogenetic and socioeconomic criteria. Twelve non-target Lepidoptera species belonging to five families were selected; these were mainly leafminers and gallformers, and included the single native Australian representative of Phyllocnistris and several weed biological control agents. Three gall-forming and one leaf-mining fly species and a single leaf-mining beetle species were also included. The outcome of host range testing indicated that the parasitoids were specific to the target species. Duan and Messing (1996, 1997) and Duan et al. (1997) studied the potential non-target impacts of Dichasmimorpha longicaudata (Ashmead), Dichasmimorpha tryoni (Cameron) and Psytallia fletcheri (Sivestri) (all Hymenoptera: Braconidae), introduced for fruit fly control in Hawaii. Two non-target species were selected based

on a suite of criteria. These included: gall size and shape or feeding niche (of the target and non-target species); relatedness of parasitoid species attacking target and nontarget hosts in the field; and shape of the parasitoid ovipositor and specialized searching behaviour. One of the non-target species studied was a native species collected in the field, and the other was a weed biological control agent that was obtained from an established culture. These studies showed that fruit shape, size and colour are essential stimuli to elicit oviposition by the candidate parasitoids and that these species would only attack fruit fly hosts that live in fruit or fruit-like structures; non-target hosts in a different feeding niche were not impacted by the biological control agents. Cameron and Walker (1997) studied the host specificity of Cotesia rubecula (Marshall) and Cotesia plutellae Kudjumov (Hymenoptera: Braconidae), parasitoids of Pieris rapae L. (Lepidoptera: Pieridae) and Plutella xylostella (L.) (Lepidoptera: Plutellidae). Initial selection of non-target species for host specificity testing was based on literature records and field collections of Lepidoptera from Brassica spp. and Urtica dioica DC in areas where C. rubecula and C. plutellae were abundant in New Zealand and Fiji. The test list was refined using behavioural data on host plant attractiveness to each parasitoid species. These studies determined that C. rubecula was highly specific to P. rapae. In contrast, despite being more attracted to species associated with cabbage volatiles, C. plutellae attacked all species tested, and successfully developed in ten of the 14 non-target species. Barratt et al. (1997, 1998, 2000) studied host specificity of two braconid parasitoids, Microctonus aethiopoides Loan and Microctonus hyperodae Loan (Hymenoptera: Braconidae), of the adult Sitona discoideus Gyllenhal and Listronotus bonariensis (Kuschel) (Coleoptera: Curculionidae), important forage pests in New Zealand. They conducted field surveys (Barratt et al., 1998) of native Curculioniodea to determine which phylogenetic, ecological and behav-


ioural affinities could be used to develop a test list. Of the 85 Curculionoidea species found, 11 were selected, and test material was collected from the field. A combination of phylogenetic and known host range information on M. aethiopoides and M. hyperodae was used to determine which non-target species would potentially be at greatest risk. Additional pest and beneficial (weed biological control agents) species related to the target species found in the surveys were included. Further criteria included similarities in feeding, seasonal abundance and activity patterns. Parasitoid behaviour patterns were also studied to determine if oviposition activities coincided with active cycles of potential non-target hosts. Laboratory results suggested that M. aethiopoides successfully developed in nine of 12, and M. hyperodae in four of 11 species tested. Field studies confirmed that M. aethiopoides parasitized a broader range of species than did M. hyperodae. Kitt and Keller (1998) carried out tests on host plant preferences of the aphid parasitoid Aphidius rosae Haliday (Hymenoptera: Aphidiidae). Results showed that only non-target aphids on roses, the habitat utilized by the target Macrosiphum rosae (L.) (Hemiptera: Aphididae), would be at risk, thereby reducing the list of non-target species to test. Species tested included those collected in sufficient numbers from glasshouses and from the field. They concluded that A. rosae would successfully attack only the target, M. rosae. Rutledge and Wiedenmann (1999) tested the response of Cotesia flavipes Cameron, Cotesia sesamiae (Cameron) and Cotesia chilonus (Matsumura), braconid parasitoids of stem-boring pests of graminaceous plants. Although no non-target testing was conducted, biological characteristics of the parasitoids and responses to a range of host and non-host plant volatiles were used as a theoretical basis for selecting non-target species. It was concluded that for certain parasitoids, testing plant preferences could help determine their ecological host range.

23

Boettner et al. (2000) studied the nontarget effects of the tachinid Compsilura concinnata (Meigen) (Diptera: Tachinidae), introduced for control of the gypsy moth Lymantria dispar (L.) (Lepidoptera: Lymantriidae), and 12 other pest species, including the saturniid moth Hemileuca oliviae Cockerell. Based on the knowledge that C. concinnata has a very broad host range (>180 native North American Lepidoptera spp.), non-target hosts selected for study were species of the same family (Saturniidae) as the target, species that feed on plants found in the same habitat (oak forest) as the main target species and those were obtainable from culture. An additional, threatened, non-target species was collected by chance in the study habitat and was incorporated into the project. These authors found that C. concinnata was responsible for significant parasitism (36% to 81%) of the three non-target species studied. Orr et al. (2000) studied host specificity of Trichogramma brassicae Bezdenko (Hymenoptera: Trichogrammatidae) and used biological and ecological criteria to determine which non-target species to include in evaluations. Based on dispersal behaviour, they determined that Lepidoptera species found in the target habitat (maize) and adjacent habitats were the most appropriate for host range testing. Furthermore, only those Lepidoptera species where eggs were present during periods of T. brassicae release were considered to be potentially vulnerable. Orr et al. (2000) collected and identified Lepidoptera species and estimated their flight period from museum collection data. Flight periods from 22 species overlapped with T. brassicae release periods, and progeny from field-collected material were used for further testing. The authors noted that species not attracted to light traps or not abundant may have been missed, particularly rare species. Of the 22 species tested in the laboratory, 11 were found to be highly suitable hosts for T. brassicae, but in the field, parasitism of these same non-target species was very low, often zero.

24

U. Kuhlmann et al.

Porter (2000) examined the host specificity of Pseudacteon curvatus Borgmeier (Diptera: Phoridae) as a biological control agent for the fire ants Solenopsis invicta Buren and Solenopsis richteri Forel (Hymenoptera: Formicidae) in the southern United States and used phylogenetic and biological information to develop a list of non-target species for testing. Information on the candidate agent indicated that only Formicidae were attacked by Pseudacteon spp., that the ovipositor of this group was highly specialized and that host size was a factor, thus limiting the ability to parasitize other organisms. Material was collected from the field for the 19 species tested. Results confirmed that P. curvatus will only develop in Solenopsis spp., and parasitism of two native species tested was considerably less than for the target species. Fuester et al. (2001) studied the host range of Aphantorhaphopsis samarensis (Villeneuve) (Diptera: Tachinidae), a candidate for biological control of gypsy moth in North America. Ecological and biological information, such as habitat and life history overlap, were considered in the selection of non-target species. Field studies in the area of origin provided information on the realized host range of A. samarensis. Of the 54 species collected in 11 families of Lepidoptera no A. samarensis emerged. Progeny of field-collected individuals (11 species from ten lepidopteran families) were primarily used in laboratory tests, although it was not stated from which habitats these species were collected. Only one non-target species, one of two Lymantriidae tested, was successfully parasitized by A. samarensis. Mansfield and Mills (2002) evaluated the host range of Trichogramma platneri Nagarkatti for control of Cydia pomonella L. (Lepidoptera: Tortricidae). They considered ecological and biological criteria (e.g. known hosts, novel hosts and host egg characteristics) to develop a list of nontarget species for testing. From this list, commercially available species and laboratory cultures that could be easily obtained were chosen for testing. The results indicated that of the 17 species tested, T. plat-

neri successfully emerged from six of 12 Lepidoptera species and the neuropteran, Chrysoperla carnea Stephens. These authors also concluded that larger eggs are generally better hosts for T. platneri. Munro and Henderson (2002) evaluated the tachinid Trigonospila brevifacies (Hardy) a parasitoid of the fruit crop tortricid Epiphyas postvittana Walker. Community-level interactions were considered when selecting non-target test species, and the list was narrowed down to species in families (Tortricidae and Oecophoridae) known to be hosts of the tachinid parasitoid. Test candidates were field-collected in the forest community. Results showed that T. brevifacies was more abundant in the field than all native parasitoids collected, and parasitized more species than did native New Zealand tachinid species. Benson et al. (2003) examined the impact of Cotesia glomerata and C. rubecula parasitoids of P. rapae on non-target Pieris spp. Phylogenetic information and ecological information were used to determine the species to be tested. The results indicated that neither of the two non-target species Pieris virginiensis Edwards and Pieris napi (Scudder), nor the target species P. rapae, were attacked in the habitat occupied by P. virginiensis. Bourchier (2003) developed a list of butterfly species that are potentially at risk if Trichogramma minutum Riley were to be mass-released in maize against Ostrinia nubilalis Hübner in Canada. Using recent taxonomic information and an existing database of 153 species he considered ecological and biological attributes (geographic distributions, oviposition, phenology, number of generations per year, overwintering stage, host-plant preferences and egg-mass type and location) to establish known host ranges of Trichogramma spp. Most species were excluded from the list because of mismatch in the geographic distributions and oviposition phenology, and some species were excluded because their biology was less known. This served as a baseline for selecting a manageable number of nontarget insects that should be subjected to host range testing. Bourchier (2003) sug-


gested that these non-target host selection criteria should be generally applied to inundative and classical biological control agents. Like Orr et al. (2000), Bourchier (2003) noted the difficulty of obtaining rare species, especially those on the ecological vulnerability list. Babendreier et al. (2003a,b) conducted laboratory and field risk assessment studies for T. brassicae using an approach similar to Bourchier (2003), considering ecological information, habitat and temporal overlap of non-target hosts and the biological control agent to select species for testing. For field tests, availability was used to determine the list of non-target species. These authors focused on butterflies as non-targets because their biology was better known, and also because butterfly biodiversity is of great concern in conservation biology. The list of 23 non-target lepidopteran species included nine species on the endangered species list in Switzerland. All species were tested in the laboratory (Babendreier et al. (2003a)) and successful parasitism was documented for 17 of the 23 species. Of the six species tested under field-cage conditions (including two species on the endangered list), all were parasitized by T. brassicae, though only at low levels. A field study with two non-target species revealed that both were parasitized at up to 2 m from the release point but parasitism at 20 m was zero. The work of Babendreier et al. (2003a,b) marks the first instance that rare butterfly species have been included in host specificity testing of biological control agents of arthropods. Babendreier et al. (2003c) studied the potential of T. brassicae to overwinter in eggs of non-target Lepidoptera. Phylogenetic information, representatives of several lepidopteran families and availability of test material were used to select the non-target species studied. T. brassicae successfully overwintered in all of the six species tested. Babendreier et al. (2003d) studied the impacts of T. brassicae on predators associated with maize. In this work non-target species were selected based on ecological

25

considerations and on availability of test material. Representative groups occurring in the target habitat were chosen and species that were commercially available were tested. Two of the four non-target predators were successfully parasitized at high levels in the laboratory, but under field conditions the levels of parasitism were very low and significantly less than for control species. In summary, a variety of strategies has been used to select species for non-target host tests. Although phylogenetic considerations were an underlying criterion (i.e. that a particular parasitoid group attacks certain host groups), ecological, biological and socio-economic information was very important for selecting non-target species for study. Availability of test material was also critical for selection of non-target test species in most studies.

Recommendations for Selecting a Species List for Host Specificity Testing using Invertebrates in Biological Control of Arthropods It is apparent that the criteria used in weed biological control are unlikely to provide all the necessary information that would enable development of a meaningful nontarget test list for entomophagous biological control agents. Arguments that have been brought forward in support of this include: ● Arthropods often outnumber plant species in communities by an order of magnitude (e.g. Kuhlmann et al., 2000; Messing, 2001). ● There is a significant lack of knowledge of arthropod phylogeny (e.g. Sands and Van Driesche, 2000; Messing, 2001). ● Natural enemies of arthropod pests respond to two trophic levels, i.e. the host and its host plant(s) (e.g. Godfray, 1994). ● Disjunct host ranges appear to be the rule with parasitoids, rather than the exception as in herbivores (Messing, 2001).

26

U. Kuhlmann et al.

● The fact that it is much more difficult and time consuming to rear a large number of test arthropod species than test plant species (Kuhlmann et al., 1998; Sands and Van Driesche, 2000). A central question with regard to the selection of test species is whether the host range of parasitoids considered for use in biological control programmes is restricted to one or a few closely related groups of herbivorous insects, or whether phylogenetic disjunction in host range, i.e. a host range that includes phylogenetically unrelated species, is the rule, rather than the exception. There seems to be consensus among arthropod biological control scientists that phylogeny is a valuable starting point for predicting and assessing the host range of parasitoids but that other criteria (e.g. ecological similarities and safeguard considerations) are also of high relevance, even more so than in host range assessment of herbivores. Thus, the selection of nontarget test species has to be carried out on a case-by-case basis. Recent studies to determine the host range of candidate entomophagous biological control agents have used an array of criteria to develop lists of species for testing the agent’s host range, as shown above in the review of methods used to date. However, there is currently no standard protocol which has been developed for test species selection. Here, we provide recommendations for developing a test list for host specificity of entomophagous arthropods (Fig. 2.1). As a first step, the information available on the recorded field hosts of the candidate biological control agent, as well as of closely related species, should be collected (see De Nardo and Hopper, 2004). Although literature reports or museum collections are important, this information should be viewed with caution, and the quality of the data assessed with a taxonomic expert. Also, it should be noted that host records tend to be primarily from agricultural and forest habitats and from economically more important species. There is general consensus that experiments are required to thoroughly determine the ecological host range of a potential biological

control agent (Hopper, 2001). Within the first step, the ecological (or realized) host range of the candidate species should be assessed through carefully planned field studies of the parasitoid–host complexes in the area of origin of the candidate biological control agent. Knowledge of the host species attacked by the candidate agent and its close relatives in the native range will facilitate the selection of appropriate test species for host range testing in the proposed area of introduction (Kuhlmann et al., 2000; Kuhlmann and Mason, 2003). In addition, comparable field studies in the area of introduction would generate valuable insight into which herbivore species would be exposed to the candidate biological control agent, both in space and time. If little is known about the target pest (see Barratt, 2004), initial studies need to be carried out to develop the information required for selection of appropriate non-target test species. Based on the knowledge of the ecological host range of the candidate biological control agent in its native range, an initial test species list should be established. We propose that this list be compiled by selecting species from three different categories, which need not be followed in any particular sequence: Category 1: Ecological Similarities: Species which live in the same/adjacent habitat (e.g. on arable land and adjacent field margins) or feed in the same microhabitat (e.g. on same plant species, or in galls) as the target species; Category 2: Phylogenetic/Taxonomic Affinities: Species which are taxonomically/phylogenetically related to the candidate biological control agent (according to modern weed biological control programmes); Category 3: Safeguard Considerations: ‘Safeguard species’, which are either beneficial insects (e.g. pollinators, other biological control agents) or rare and endangered species that belong to the same family or order. Additionally, host species of congeneric species of the candidate biological control agent could be selected when appropriate.


27

Ecological Host Range Information

Category 1: Ecological Similarities

Category 2: Phylogenetic/ Taxonomic Affinities

Category 3: Safeguard Considerations

Initial Test List

Filter 1: Spatial, Temporal and Morphological Attributes

Filter 2: Accessibility and Availability

Revised Test List

New Information

Host Specificity Testing

Fig. 2.1. Recommendations for the selection of non-target species for a test list to be applied in host specificity testing of invertebrates for biological control of arthropods.

Depending on the information available, one may prioritize in the test list either species related to the target host or species that feed in the same microhabitat. Priority should be given to selecting species that are associated with more than one category. The initial list may consist of 50 or more test species and may be comparable to the final test plant list in a weed biological control programme. However, it is much more laborious and time-consuming to rear 50 or more insect species than it is

to grow plant species. Collection of suitable stages of test species may be possible, but requires evidence that the collected stages are not already parasitized or diseased. Holding field-collected individuals for non-target testing in a laboratory colony is recommended to ensure that any field parasitism or natural disease runs its course. Sands (1997) stated that testing of more than ten species of non-target arthropods may be impractical and often unnecessary. Further, Sands (1998) suggested

28

U. Kuhlmann et al.

that carefully designed tests on a few species related to the target will provide adequate information relating to the host specificity of candidate agents. We therefore propose reduction of the list of test species by filtering out those species with attributes which do not overlap with those of the target species. Attributes that may lead to a species being discarded from the test list are non-overlapping geographical distribution, different climate requirements, phenological asynchronization or host size which is outside the range that is accepted by the candidate biological control agent (Filter 1 in Fig. 2.1). The latter attribute can be tested by offering target species or other host species of different size classes to the candidate biological control agent. Other attributes may be investigated by studying the herbivore complex that inhabits the area into which the candidate biological control agent is to be released. Some of the species remaining in the test list are not available or accessible in large enough numbers and they should not be considered for host specificity testing as an adequate number of replicates cannot be conducted (Filter 2 in Fig. 2.1). In the case of rare and endangered species consideration can be given to testing congenors as surrogates. The revised test species list may then include some ten to 20 test species. Although this revised list would be appropriate for starting host specificity testing, it should not necessarily be considered as the final test list. Results from ongoing host specificity testing and parallel studies that aim to assess the chemical, visual and tactile cues emitted by the host or its hostplant(s), and involved in the agent’s host-selection behaviour, may shed new light on which non-target species may be at risk of being attacked by the candidate biological control agent. We therefore propose that the revised test species list should be periodically revisited during the prerelease studies of arthropod biological control programmes (indicated by the feedback loop in Figure 2.1). New information gath-

ered during the pre-release studies may lead to scientifically based justification for removal or addition of test species. Such a scenario is also applied in weed biological control programmes. In North America, test plant lists that have been submitted to and approved by the Technical Advisory Group at the beginning of a programme may be subject to well-founded revision during later stages of the pre-release studies. However, we believe that this reiterative process is of greater relevance in arthropod biological control programmes because of the need to keep the test list as short as possible, while still providing a reliable host range profile for the candidate biological control agent.

Selection of Non-target Species for a Test List: a Case Study Celatoria compressa, an adult parasitoid of species in the subtribe Diabroticina in North America, was selected as a candidate for classical biological control of Diabrotica virgifera virgifera (Coleoptera: Chrysomelidae: Galerucinae) in Europe. Prior to its potential release, host specificity testing was conducted to evaluate the potential impacts of C. compressa on European indigenous Coleoptera species. The non-target species selection recommendations described above were applied to select appropriate indigenous non-target species for host specificity testing of C. compressa under quarantine conditions.

Ecological host range information Information was compiled about the known field host ranges of Celatoria species, such as C. compressa, C. bosqi Blanchard, C. diabroticae (Shimer) and C. setosa (Coquillet), based on published host–parasitoid rearing records from North, Central and South America. Based on literature records, the known ecological host range of the three betterknown Celatoria species (C. bosqi, C. diabroticae and C. setosa) is restricted to the


subtribe Diabroticina within the tribe Luperini of the subfamily Galerucinae. Celatoria bosqi, present in South America, is known to parasitize Diabrotica speciosa (Germar) (Blanchard, 1937; Heineck-Leonel and Salles, 1997), D. sp. nr. fulvofasciata Jacoby and D. viridula (F.) (G. Cabrera Walsh, Buenos Aires, 2003, personal communication) and the chrysomelid Cerotoma arcuata Olivier (Magalhães and Quintela, 1987). The ecological host range from the North American C. diabroticae is restricted to D. undecimpunctata howardi Barber, D. undecimpunctata undecimpunctata Mannerheim, D. longicornis (Say) and D. v. virgifera (Fisher, 1983). Although recorded hosts of the North American C. setosa include Diabrotica species (Arnaud, 1978), field and experimental data indicated that it was almost exclusively a parasitoid of Acalymma species, such as the chrysomelids Acalymma blandula LeConte, A. trivittata (Mannerheim) and A. vittata (F.) (Fischer, 1981, 1983). As ecological host range information about C. compressa was mostly not available, field host range surveys were carried out in Mexico, the area of origin. Celatoria compressa was found to only parasitize D. v. virgifera, D. balteata LeConte, D. porracea Harold, D. scutellata Baly, D. tibialis Baly, D. viridula, Acalymma blomorum Munroe and Smith, A. fairmairei (F.), A. innubum (F.), A. trivittata, Gynandrobrotica spp. and Cerotoma atrofasciata Jacoby (Eben and Barbercheck, 1996; A. Eben, Xalapa, Mexico, 2003, personal communication). The information on the ecological host range provided evidence that C. bosqi, C. diabroticae and C. setosa, as well as C. compressa, are highly specialized. Thus, most probably all species in the genus Celatoria parasitize only adults of single or related genera within the subfamily Galerucinae (most probably at the tribe level of Luperini), or Alticinae in the family Chrysomelidae (see Cox, 1994 and publications mentioned above). Additionally, it was reported by Fischer (1983) for C. diabroticae and C. setosa, and by Zhang et al. (2003) for C. compressa, that females use a

29

piercing ovipositor to successfully parasitize hosts. Therefore, it is likely that Celatoria species have a high degree of host specificity compared to many other tachinids due to the elaborately modified piercing ovipositor (Belshaw, 1994; J. O’Hara, Ottawa, 2000, personal communication). Based on these findings, the selection of non-target coleopteran species for testing should be limited to the family Chrysomelidae. Ecological similarities (Category 1) Literature records were used to compile a list of the Coleoptera species which occur in selected European agricultural habitats such as maize (Zea mays L.), lucerne (Medicago sativa L.), pumpkin (Cucurbita maxima Duch.), wheat (Triticum aestivum L.) and sunflower (Helianthus annuus L.), as well as in adjacent field margin habitats. These habitats were selected because they are commonly present in the area invaded by the target. A total of 185 coleopteran species (belonging to 14 families) were found to be associated with these selected habitats in Europe. From these 185 species, three Galerucinae, 22 Alticinae, six Chrysomelinae, five Criocerina and two Cassidinae species (all in the family Chrysomelidae) were included (in total, 38 species for the initial test list). With regard to species living in the same microhabitat (same host plant) no obvious candidates were found but the cereal leaf beetle, Oulema melanopus (L.), which occasionally feeds in the same microhabitat (maize), and has been considered in the selection process. Phylogenetic/Taxonomic affinities (Category 2) The phylogenetic relationship of the nontarget species to the target was checked to ensure that European species of related subfamily, genera or subtribes of the target were added to the non-target list. In addition, a representative non-target species from a genus in a different family within the same order (outgroup) was selected.

30

U. Kuhlmann et al.

As reported before, D. v. virgifera belongs to the tribe Luperini (subtribe Diabroticina) within the subfamily Galerucinae (Wilcox, 1972), therefore, representative species closely related to D. v. virgifera were considered. Further, phylogenetic studies by Hsiao (1994) have shown that the subfamily Galerucinae is closely related to the subfamilies Criocerinae, Chrysomelinae and Alticinae, and relatively distant from Cassidinae. In this case study, representative non-target species selected belonged to the subfamilies Criocerinae (e.g. O. melanopus), Chrysomelinae (e.g. Gastrophysa viridula Deg. and Gonioctena fornicata Brüggemann) and Cassidinae (e.g. Cassida rubiginosa Müller). Within the subfamily Galerucinae, other representative non-target species in the tribe Galerucini, such as Galerucella pusilla Duft and Pyrrhalta luteola (Müller), were chosen. In addition, a species in the tribe Luperini, Aulacophora foveicollis Lucas (subtribe Aulacophorinia), which represents a species of the genus Diabrotica in the Old World (Maulik, 1936), was selected. Besides this chrysomelid, the pea weevil, Sitona lineatus L., was selected as the outgroup representative of a different and not closely related Coleoptera family (Coleoptera: Curculionidae); this is a common species present in Diabrotica-invaded areas (in total, two additional species for the initial test list).

Safeguard considerations (Category 3) Representatives of beneficial insect families such as Coccinellidae or Carabidae, as well as weed biological control agents, were included to avoid non-target impacts on these organisms. In addition, rare and endangered species were considered for selection. The two-spotted ladybird beetle, Adalia bipunctata L., was added to the non-target list as a representative of beneficial Coccinellinae (Coleoptera: Coccinellidae), and the golden loosestrife beetle, G. pusilla, considered as an important species for the control of the weed, purple loosestrife (Lythrum salicariae L.) in Europe (two additional species for the initial test list).

Initial test list Taking the above results, a total of 42 species were included in the initial test list (38 (category 1) + 2 (category 2) + 2 (category 3) = 42 species). It should be noted that G. pusilla was selected under both categories 2 and 3, and O. melanopus was selected under categories 1 and 2, illustrating that species can fulfil multiple information requirements. Spatial, temporal and morphological attributes (Filter 1) The initial test list was progressively filtered (reduced), due to the fact that nontarget species potentially at risk need to have ecological and biological attributes that may or may not overlap with those of the target species. In this case, geographical distribution and climate requirements (European continent excluding UK and Scandinavia), temporal pattern of adult occurrence in the field (June till October) and similarity in size (3–10 mm required for parasitoid development within the host; Eben and Barbercheck, 1996) were used. As a result of Filter 1, 21 chrysomelid species were excluded due to body size (>10 mm and 1.2 ha) with insecticide baits. They compared densities of arthropod predators throughout the growing season in treated vs control plots, and were able to quantify the predators’ susceptibility to predation by the fire ant. Colfer et al. (2003) recently used a similar approach to examine the role of naturally occurring generalist arthropod predators on the establishment and efficacy of a predatory mite, Galendromus occidentalis (Nesbitt) (Acarina: Phytoseiidae), that is commonly released for the control of the spider mite Tetranychus urticae Koch (Acarina: Tetranychidae). They first used small field cages to test interactions with different combinations of predators over short-term periods in cotton fields. Next, they employed insecticide manipulations to examine long-term interactions on a larger spatial scale. They concluded that hemipteran predators have a negative impact on predatory mite populations via intra-guild predation. However, generalist predators did not curtail the long-term biological control of spider mites, because their addition to the T. urticae–G. occidentalis system caused a significant overall reduction in pest densities. To a large extent the methods referred to above cannot be used in the target country prior to release of a particular natural enemy in question. They can, however, be

Measuring and Predicting Indirect Impacts of Biological Control

of great value when conducted post-release to help establish patterns and principles that guide overall analyses of indirect impact and risk assessment. For individual arthropod introductions it may be possible to conduct similar studies in the country of origin, prior to collection and importation. In one example, Liu et al. (2004) obtained baseline information on species interactions with natural enemies of the soybean aphid, Aphis glycines Matsumura (Hemiptera: Aphidae), in China, prior to their importation to North America. While not giving the kinds of detailed information alluded to above (because some of the species from the target country will not be present in the country of origin), nevertheless some basic patterns may be discerned. Biological control projects are practical experiments in applied ecology, and the extent of pre-release testing is invariably limited by logistical considerations such as funding, time constraints, manpower and political realities. In many cases it will be logistically or economically impossible to conduct the extent of research desired in the country of origin. In recent explorations for Mediterranean fruit fly parasitoids in Kenya, for example, our research efforts were pummelled by floods, impassable roads, worker illness and lethal terrorist attacks. Nevertheless, even when logistical considerations preclude complex manipulative experiments, simple surveys of diet breadth and habitat use in the country of origin can provide invaluable information that should not be overlooked.

Surrogate experiments Even with the seemingly intractable problems associated with predicting competitive interactions prior to the release of a new biological control agent into a complex ecosystem, it may be possible in some situations to use surrogate species in field experiments. These types of experiments have great potential after theoretical explorations have identified critical features of the interaction, and subsequent analyses demonstrate that the surrogate is a biologically

73

close and realistic stand-in for the natural enemy in question. For example, substantial differences in the intrinsic rate of increase between the surrogate and focal species could render such experiments moot. In an elegant field demonstration of indirect impacts, van Nouhuys and Hanski (2000) showed how apparent competition mediated by a generalist hyperparasitoid might be predicted using a surrogate primary parasitoid. The specialist braconid wasp Cotesia melitaearum (Wilkinson) (Ichneumonoidae: Braconidae), which attacks the Granville fritillary butterfly Melitaea cinxia (Linnaeus) (Lepidoptera: Nymphalidae) on islands near Finland, is itself attacked by the generalist ichneumonid hyperparasitoid Gelis agilis (Fabricius) (Ichneumonoidae: Cryptinae). The researchers experimentally added to the system a second braconid host of G. agilis (Cotesia glomerata (L.)), which does not compete with C. melitaearum for resources (thus controlling for any effects of exploitation competition). Their replicated experiments in the field were able convincingly to demonstrate that the addition of the new primary parasitoid increased extant populations of the hyperparasitoid, which subsequently reduced populations of C. melitaearum in all replicates (so much so, in fact, that two of the original populations were driven to extinction). One can envisage how, in a similar manner, surrogates could be used to test for possible indirect impacts of introducing a new parasitoid in addition to an existing, partially effective, parasitoid in a biological control programme. The key, of course, is having an available surrogate species that already exists in the target region (i.e. does not itself have to go through quarantine), and one that is biologically similar to the prospective import. While it is unlikely that the risk of apparent competition could be quantified definitively using this method, it could add a valuable dimension to a comprehensive risk analysis, and offer guidelines for post-release follow-up studies should release of the new parasitoid be approved.

74

R. Messing et al.

Rules of Thumb Biological control risk analysis is a very new discipline, and is still struggling to come to terms with methods for measuring even direct impacts of predators and parasitoids. For subtle indirect effects, there is still much that needs to be learned in order to improve, integrate and make best use of the methods discussed above. In the near future the choice, outcome and interpretation of these types of tests will necessarily have to be integrated with common sense and our best current understanding of parasitoid (or predator) ecology and community dynamics. The current biological control-permitting system in most countries makes use, implicitly, of expert opinion in providing guidance to decision-makers. It may be advantageous to make this approach explicit, and to formalize the ‘expert systems’ approach so as to take fullest advantage of our current knowledge, imperfect as it may be. There is ample precedent for using expert systems in other aspects of integrated pest management (e.g. Messing et al., 1989). In very general terms there may be some ‘rules of thumb’ based on arthropod life history parameters that can help decision-makers generalize to some extent about the risk of particular importations. Most obvious is the breadth of host (or prey) range (see van Lenteren et al., Chapter 3, this volume); highly polyphagous species will generally be riskier than monophagous or stenophagous ones. Hymenoptera that can act as facultative hyperparasitoids are more likely to have indirect effects than are obligate primary parasitoids (Brodeur, 2000). Some parasitoid taxa are known to be more vulnerable to hyperparasitism; these indicate an increased risk for apparent competition with native parasitoids (Heimpel et al., 2004). While bearing in mind these considerations, however, one must guard against oversimplification, and recognize that the idiosyncratic nature of extremely diverse parasitoid life

histories will always require detailed analyses of individual species and even of sub-species genetic cohorts.

Conclusions All of the techniques that have been mentioned in this chapter have their own advantages and their own limitations. None of them, in themselves, are able to predict accurately the full extent of competitive and other indirect interactions that will follow upon the introduction of a new species into an existing community matrix. Natural ecosystems, and even simplified agricultural environments, are usually too complex and their relationships too subtle for us to know in advance to what extent new species will alter existing patterns. Nevertheless, as in many fields of human endeavour, decisions must be made even though our knowledge is imperfect. Biological control is but one option for managing invasive species, and these invasives often have significant negative environmental and economic impacts. Alternative choices for pest management, including the choice to do nothing, have their own risks and also proceed with imperfect knowledge. In order to minimize risk when introducing new biological control agents, we suggest that a comprehensive overview of the organism’s role in the ecological community be outlined, using a combination of the techniques mentioned here along with a thorough literature evaluation of the species and its congeners in the area of origin. We also argue that a sustained and well-funded effort to retrospectively evaluate the case histories from prior biological control programmes would help shore up the empirical data base upon which theory and modelling can build a broader picture of community dynamics and the response to insertion of new species. While risk cannot be eliminated, it can be managed with increasing confidence as our understanding of community dynamics grows incrementally.


Acknowledgements This work was supported in part by USDAARS grant No. 5853208147 to RHM and

75

NSERC Operating grants to BDR and JB. We also thank George Heimpel and an anonymous reviewer for constructive comments on an earlier version of this chapter.

References Abrams, P. (2004) Trait-initiated indirect effects due to changes in consumption rates in simple food webs. Ecology 85, 1029–1038. Abrams, P.A. and Matsuda, H. (2003) Population dynamical consequences of reduced predator switching at low total prey densities. Population Ecology 45, 175–185. Asquith, A. and Miramontes, E. (2001) Alien parasitoids in native rainforests: the ichneumonoid wasp community in a Hawaiian rain forest. In: Lockwood, J., Howarth, F. and Purcell, M. (eds) Balancing Nature: Assessing the Impact of Importing Non-Native Biological Control Agents. Thomas Say Publications in Entomology, ESA, Lanham, Maryland, pp. 54–69. Bokonon-Ganta, A.H., Ramadan, M.M., Wang, X.G. and Messing, R.H. (2005) Biological performance and potential of Fopius ceratitivorus (Hymenoptera: Braconidae), an egg-pupal parasitoid of tephritid fruit flies, newly imported to Hawaii. Biological Control 33, 238–247. Borer, E.T., Murdoch, W.W. and Swarbrick, S.L. (2004) Parasitoid coexistence: linking spatial field patterns with mechanism. Ecology 85, 667–678. Brodeur, J. (2000) Host specificity and trophic relationships of hyperparasitoids. In: Hochberg, M.E. and Ives, A.R. (eds) Parasitoid Population Biology. Princeton University Press, Princeton, New Jersey, pp. 163–183. Casas, J. (2000) Host location and selection in the field. In: Hochberg, M.E. and Ives, A.R. (eds) Parasitoid Population Biology. Princeton University Press, Princeton, New Jersey, pp. 17–26. Cisneros, J.J. and Rosenheim, J.A. (1998) Changes in the foraging behavior, within-plant vertical distribution and micro-habitat selection of a generalist predator: an age analysis. Environmental Entomology 27, 949–957. Colfer, R.G. and Rosenheim, J.A. (2000) Predation on immature parasitoids and its influence on aphid population suppression. Oecologia 126, 292–304. Colfer, R.G., Rosenheim, J.A., Godfrey, L.D. and Hsu, C.L. (2003) Interactions between the augmentatively released predaceous mite Galendromus occidentalis (Acari: Phytoseiidae) and naturally occurring generalist predators. Environmental Entomology 32, 840–852. Collier, T., Kelly, S. and Hunter, M. (2002) Egg size, intrinsic competition, and lethal interference in the parasitoids Encarsia pergandiella and Encarsia formosa. Biological Control 23, 254–261. Connell, J.H. (1980) Diversity and the coevolution of competitors, or the ghost of competition past. Oikos 35, 131–138. Connell, J.H. (1983) On the prevalence and relative importance of interspecific competition: evidence from field experiments. American Naturalist 122, 661–669. Cooper, G. (1993) The competition controversy in ecology. Biology and Philosophy 8, 359–384. Coulson, J.R., Soper, R.S. and Williams, D.W. (1991) Biological Control: Needs and Procedures. Proceedings Workshop, US Department of Agriculture, Agricultural Research Service, Washington DC. DeBach, P. and Rosen, D. (1991) Biological Control by Natural Enemies. Cambridge University Press, Cambridge, UK. Eubanks, M.D., Blackwell, S.A., Parrish, C.J., Delamar, Z.D. and Hull-Sanders, H. (2002) Intraguild predation of beneficial arthropods by red imported fire ants in cotton. Environmental Entomology 31, 1168–1174. Follett, P.A. and Duan, J.J. (2000). Nontarget Effects of Biological Control. Kluwer, Norwell, Massachusetts. Greathead, D.J. (1995) Benefits and risks of classical biological control. In: Hokkanen, H. and Lynch, J. (eds) Biological Control: Risks and Benefits. Cambridge University Press, Cambridge, UK, pp. 53–63. Heimpel, G.E., Rosenheim, J.A. and Mangel, M. (1997) Predation on adult Aphytis parasitoids in the field. Oecologia 110, 346–352.

76

R. Messing et al.

Heimpel, G.E., Ragsdale, D.W., Venette, R.C., Hopper, K.R., O’Neil, R.J., Rutledge, C.E. and Wu, Z. (2004) Prospects for importation biological control of the soybean aphid: anticipating potential costs and benefits. Annals of the Entomological Society of America 97, 249–258. Holt, R.D. (1995) Community modules. In: Gange, A.C. and Brown, V.K. (eds) Multitrophic Interactions in Terrestrial Systems. Blackwell Science, Oxford, UK, pp. 333–350. Holt, R.D. and Hochberg, M.E. (2001) Indirect interactions, community modules and biological control: a theoretical perspective. In: Wajnberg, E., Scott, J.K. and Quimby, P.C. (eds) Evaluating Indirect Ecological Effects of Biological Control. CABI Publishing, Wallingford, UK, pp. 13–37. Hoogendoorn, M. and Heimpel, G.E. (2001) PCR-based gut content analysis of insect predators: using ribosomal ITS-1 fragments from prey to estimate predation frequency. Molecular Ecology 10, 2059–2067. Hoogendoorn, M. and Heimpel, G.E. (2004) Competitive interactions between an exotic and a native ladybeetle: a field cage study. Entomologia Experimentalis et Applicata 111, 19–28. Howarth, F.G. (1990) Environmental impacts of classical biological control. Annual Review of Entomology 36, 485–509. Howarth, F.G. (1991) Hawaiian terrestrial arthropods: an overview. Bishop Museum Occasional Papers 30, 4–26. Jervis, M.A. and Kidd, N.A.C. (1996) Insect Natural Enemies. Chapman and Hall, London. Kareiva, P. and Odell, G. (1987) Swarms of predators exhibit ‘preytaxis’ if individual predators use area-restricted search. American Naturalist 130, 233–270. Lima, S.L. (1998) Stress and decision making under the risk of predation: recent developments from behavioral, reproductive, and ecological perspectives. Advances in the Study of Behavior 27, 215–290. Liu, J., Wu, K., Hopper, K.R. and Zhao, K. (2004) Population dynamics of Aphis glycines (Homoptera: Aphididae) and its natural enemies in soybean in Northern China. Annals of the Entomological Society of America 97, 235–239. Lockwood, J.A., Howarth, F.G. and Purcell, M.F. (2001) Balancing Nature: Assessing the Impact of Importing Non-Native Biological Control Agents (An International Perspective). Thomas Say Publications in Entomology, ESA, Lanham, Maryland. Lonsdale, W., Briese, D. and Cullen, J. (2001) Risk analysis and weed biological control. In: Wajnberg, E., Scott, J. and Quimby, P. (eds) Evaluating Indirect Ecological Effects of Biological Control. CABI Publishing, Wallingford, UK, pp. 185–210. Louda, S., Pemberton, R., Johnson, M. and Follett, P.A. (2002) Non-target effects – the Achilles heel of biological control? Retrospective analyses to reduce risk associated with biocontrol introductions. Annual Review of Entomology 48, 365–396. Lucas, E., Coderre, D. and Brodeur, J. (1998) Intraguild predation among aphid predators: characterization and influence of extraguild prey density. Ecology 79, 1084–1092. May, R.M., Hassell, M.P., Anderson, R.M. and Tonkyn, D.W. (1981) Density dependence in host-parasitoid models. Journal of Animal Ecology 50, 855–865. McEvoy, P. and Coombs, E. (2000) Why things bite back: unintended consequences of weed biological control. In: Follett, P.A. and Duan, J.J. (eds) Nontarget Effects of Biological Control. Kluwer, Norwell, Massachusetts, pp. 167–194. Messing, R.H., Croft, B.A. and Currans, K. (1989) Using expert system technology in natural resource management. Applications in Natural Resource Management 3, 1–11. Ostman, O. and Ives, A.R. (2003) Scale-dependent indirect interactions between two prey species through a shared predator. Oikos 102, 505–514. Peacor, S.D. (2002) Positive effect of predators on prey growth rate through induced modifications of prey behaviour. Ecology Letters 5, 77–85. Pearson, D.E. and Callaway, R.M. (2003) Indirect effects of host-specific biological control agents. Trends in Ecology and Evolution 18, 456–461. Pérez-Lachaud, G., Batchelor, T.P. and Hardy, I.C.W. (2004) Wasp eats wasp: facultative hyperparasitism and intra-guild predation by bethylid wasps. Biological Control 30, 149–155. Powell, W., Walton, M.P. and Jervis, M.A. (1996) Populations and communities. In: Jervis, M. and Kidd, N. (eds) Insect Natural Enemies. Chapman and Hall, London, pp. 223–292. Roitberg, B. (2000) Threats, flies and protocol gapes: can behavioral ecology save biological control? In: Ives, A. and Hochberg, M. (eds) Parasite Population Biology. Princeton University Press, Princeton, New Jersey, pp. 254–265.


77

Rosenheim, J.A. (2005) Intraguild predation of Orius tristicolor by Geocoris spp.; the paradox of irruptive spider mite dynamics in California cotton. Biological Control 32, 172–179. Rosenheim, J.A., Wilhoit, L.R. and Armer, C.A. (1993) Influence of intraguild predation among generalist insect predators on the suppression of an herbivore population. Oecologia 96, 439–449. Rosenheim, J.A., Kaya, H.K., Ehler, L.E., Marois, J.J. and Jaffee, B.A. (1995) Intraguild predation among biological-control agents – theory and evidence. Biological Control 5, 303–335. Schmitz, O.J., Beckerman, A.P. and O’Brien, K.M. (1997) Behaviorally mediated trophic cascades: Effects of predation risk on food web interactions. Ecology 78, 1388–1399. Simberloff, D. and Stiling, P. (1996) How risky is biological control? Ecology 77, 1965–1974. Stiling, P. and Simberloff, D. (2000) The frequency and strength of non-target effects of invertebrate biological control agents of plant pests and weeds. In: Follett, P.A. and Duan, J.J. (eds) Nontarget Effects of Biological Control. Kluwer, Norwell, Massachusetts, pp. 31–44. Symondson, W.O.C. (2002) Molecular identification of prey in predator diets. Molecular Ecology 11, 627–641. van Nouhuys, S. and Hanski, I. (2000) Apparent competition between parasitoids mediated by a shared hyperparasitoid. Ecology Letters 3, 82–84. Wajnberg, E., Scott, J. and Quimby, P. (2001) Evaluating Indirect Ecological Effects of Biological Control. CABI Publishing, New York. Wang, X.G. and Messing, R.H. (2002) Newly imported larval parasitoids pose minimal competitive risk to extant egg-larval parasitoid of fruit flies in Hawaii. Bulletin of Entomological Research 92, 423–429. Wang, X.G. and Messing, R.H. (2003) Intra- and interspecific competition by Fopius arisanus and Diachasmimorpha tryoni (Hymenoptera: Braconidae), parasitoids of Mediterranean fruit fly Ceratitis capitata (Diptera: Tephritidae). Biological Control 27, 251–259. Wang, X.G. and Messing, R.H. (2004a) Potential interactions among pupal and egg- or larval-pupal parasitoids of tephritid fruit flies. Environmental Entomology 33, 1313–1320. Wang, X.G. and Messing, R.H. (2004b) Two different life-history strategies determine the competitive outcome between Dirhinus giffardii (Chalcididae) and Pachycrepoideus vindemmiae (Pteromalidae), ectoparasitoids of cyclorrhaphous Diptera. Bulletin of Entomological Research 94, 473–480.

5

Risks of Interbreeding Between Species Used in Biological Control and Native Species, and Methods for Evaluating Their Occurrence and Impact Keith R. Hopper,1 Seth C. Britch1 and Eric Wajnberg2 1USDA,

ARS, Beneficial Insects Introduction Research Laboratory, 501 South Chapel Street, Newark, Delaware 19713, USA (email: [email protected]; [email protected]; fax number: +1-302-737-6780); 2INRA, 400 Route des Chappes, BP 167, 06903 Sophia-Antipolis Cedex, France (email: [email protected]; fax number: +33-4-92-38-6557)

Abstract Insect species introduced or augmented for biological control of insect pests may interbreed with native species, which may change fitness or cause evolution, which may in turn alter abundances. By ‘interbreeding’, we mean any reproductive interactions between species. We review the literature on factors affecting the likelihood of interbreeding between insect species and the impacts when these occur. We discuss phylogenetic relatedness, geographical distribution, spatial and temporal barriers to mating, mate recognition, copulation and sperm use, hybrid inviability and sterility, hybrid speciation, reproductive character displacement and introgression. We concentrate on the risks from introduced species, but we also address the risks from augmentation of native species. We propose methods for pre-introduction or pre-augmentation assessment of the likelihood and potential impact of interbreeding between native species and insects used in biological control. Finally, we propose methods for evaluating the occurrence and impact of interbreeding after insect species are introduced or augmented.

Introduction Exotic species introduced into a new region may court, mate, hybridize or introgress with native species, and these interactions may change fitness (Oliver, 1979; Rawlings, 1985; Presgraves, 2002) or cause evolution (Ewel et al., 1999; Cox, 2004), which may in turn alter abundances 78

(Simberloff and Stiling, 1996). Native species augmented in abundance for biological control may also court, mate or hybridize with other native species, which may cause changes in their fitness and alter their abundances (Pinto et al., 2003). By ‘court’, we mean recognize one another as potential mates, but not necessarily sufficiently to copulate. By ‘mate’ we mean


Risks of Interbreeding Between Species

copulation, with or without sperm transfer, which may or may not produce progeny. By ‘hybridize’, we mean produce hybrid progeny, which may or may not be viable or fertile. By ‘introgress’, we mean transfer of DNA sequences between species, which may or may not affect fitness, behaviour or ecology, and may or may not persist and spread in the receiving species. By ‘interbreeding’, we mean any or all such reproductive interactions between species. The risks of interbreeding apply to intentional introductions, like those in the horticultural and pet trades (Frank and McCoy, 1995; Young et al., 1999; Walker et al., 2002), as well as to accidental introductions (Yukawa, 1996; Sagarra and Peterkin, 1999; Swanson et al., 2000). Insect species introduced for biological control of insect pests may interbreed with native species (Huxel, 1999; Mooney and Cleland, 2001), although we have found few studies on such interbreeding, and only one potential case of hybridization (Yara et al., 2000). Here we review the literature on factors affecting the likelihood of interbreeding between insect species and the impacts when these occur. We concentrate on the risks from introduced species, but we also discuss the risks from augmentation of native species. We propose methods for pre-introduction or pre-augmentation assessment of the likelihood and potential impact of interbreeding between native species and insects used in biological control. Finally, we propose methods for evaluating the occurrence and impact of interbreeding after insect species have been introduced or augmented. We deal primarily with interbreeding among species, rather than groups below the level of species, because we assume that introductions of subspecies or populations of species that already occur in the target region are rare except for multiple introductions of exotic agents. Distinguishing closely related species can be difficult, particularly among parasitic Hymenoptera (Davis et al., 1987; Danforth et al., 1998; Hoy et al., 2000). Indeed, taxa described as species have been found to be complexes of sibling species which

79

differ in traits like host specificity (Davis et al., 1987; Nyman, 2002). Native sibling species are unlikely to be at risk from biological control introductions because candidates for introduction would not be considered if what appeared to be the same species already occurred in the target region. However, native species might be at risk from interbreeding with sibling species augmented for biological control. Because data on interbreeding as a result of biological control introductions or augmentations are rare, we draw on the broader literature about interbreeding among insect species, while concentrating on evidence from the orders most frequently used to control insect pests: Diptera and Hymenoptera (Clausen, 1978). Research on species in the genus Trichogramma (Hymenoptera: Trichogrammatidae) provides examples on how to measure several attributes important in risk assessment of interbreeding. Thus, we use this research as a case history throughout the paper. Although we cover published literature in our review, useful data (e.g. concerning geographic distribution and habitats and hosts used) for particular projects can be obtained from museum collections and other unpublished sources (project reports, quarantine records). Encounters between introduced and native insect species in biological control are necessarily contacts between newly sympatric species that were previously allopatric. Thus, the literature on interbreeding between allopatric species after secondary contact (Mayr, 1963; Dobzhansky, 1970; Kohlmann and Shaw, 1991; Brennan and Fairbairn, 1995; Shoemaker et al., 1996; Sperling et al., 1996; Willett et al., 1997) is directly relevant. However, we do not mean to say that introduced and native species arose by allopatric speciation. Indeed, native versus introduced species considered in biological control could have arisen by any of the proposed mechanisms of speciation: sympatric, allopatric, peripatric or parapatric; mediated by ecological selection, sexual selection, allopolyploidy, genetic drift or symbiotes. The literature on hybrid zones (Endler

80

K.R. Hopper et al.

1977; Moore, 1977; Barton and Hewitt, 1981, 1985; Harrison, 1986; Howard, 1986; Barton and Gale, 1993) is relevant to interbreeding between native species and insects introduced or augmented for biological control. For introduced insects, the appropriate hybrid zone model is likely to change with time after introduction. Initially, the distribution of introduced populations may overlap completely with native species at risk of interbreeding, but native species may have large regions of allopatry. This is very different from the usual models of hybrid zones where both species have regions of allopatry with a more or less narrow region of sympatry where hybridization can occur. Further along after introduction, introduced species may spread throughout the region where target pests occur and may broadly overlap the distribution of native species. On the other hand, introduced species may not initially overlap distributions of some native species and may only come into contact with them after some period of spread. The width and stability of hybrid zones will depend on dispersal, habitat specificity and the relative fitness of hybrids versus parental populations. For augmented native species, the distribution of hybrid zones will depend on the spatial distribution of releases and on dispersal rates.

edness and the likelihood or impact of interbreeding varies among taxa and depends as well on whether species are allopatric versus sympatric, or ecologically and behaviourally similar versus dissimilar (Coyne and Orr, 1997). In insects, all reported cases of hybridization are among species in a single genus or in a species complex within a genus (Table 5.1). Lack of observations of interbreeding among more distantly related species may arise from a bias towards searching for such interactions only where one expects to find them. Nonetheless, effort on interbreeding in biological control should concentrate on closely related species in the same complex or genus. A centrifugal approach, like that used in host range testing, may prove useful (see van Lenteren et al., Chapter 3, this volume). Case history Pinto et al. (1992) examined molecular genetic differences among 22 cultures of the Trichogramma minutum Riley complex and established phylogenetic groups that strongly predicted reproductive compatibility. In contrast, Pinto et al. (1991) showed that taxonomic grouping of T. pretiosum Riley, T. deion Pinto and Oatman and T. minutum, based on morphology, correlated poorly with reproductive compatibility, even among populations considered conspecific.

Factors Affecting Interbreeding Geographic distribution Phylogenetic relatedness Species that are phylogenetically close are more likely to interbreed than species that are phylogenetically distant (Coyne and Orr, 1997). Phylogenetic relatedness can be determined using molecular, behavioural and morphological data. Thus, one could perhaps delimit native species at risk for interbreeding with introduced species based on phylogenetic proximity. However, for many taxa, we have no phylogenies, only taxonomic keys based on morphology and only loosely related to phylogeny. Furthermore, the relationship between relat-

Climatic, habitat and geographic barriers to the spread of the introduced species may prevent sympatry with some native species, even after introduction. Thus, whether a particular native species is at risk for interbreeding depends on its distribution, the climatic tolerances of the introduced species and the ability of the introduced species to disperse across habitat and geographic barriers like grasslands, deserts and mountains. For augmentative releases, the geographic distribution of augmentation programmes will determine which native species are at risk of interbreeding.

Table 5.1. Studies showing hybrids between insect species in orders with taxa used for biological control. Family

Species

Diptera

Chironomidae Chironomidae

Drosophilidae Tephritidae Tephritidae Tephritidae Tephritidae

Glyptotendipes pallens Meigen, G. glaucus Meigen Chironomus usenicus Loginiva and Belyanina (C. plumosus Linnaeus ⫻ C. behningi Goetghebuer) Anopheles gambiae Giles, A. arabiensis Patton Culex pipiens Linnaeus, C. quinquefasciatus Say Anopheles hyrcanus Pallas group Aedes triseriatus Say, A. brelandi Zavortink, A. hendersoni Cockerell Aedes triseriatus Say, A. zoosophus Dyar and Knab Anopheles bwambae White, A. gambiae Giles Aedes scutellaris Walker group (A. cooki Belkin ⫻ A. kesseli Huang and Hitchcock) Drosophila spp. Eurosta solidaginis Fitch Anastrepha sororcula Zucchi, A. obliqua Macquart Bactrocera tryoni Froggatt, B. neohumeralis Hardy Anastrepha fraterculus Wiedemann complex

Aphelinidae Apidae Cynipidae Formicidae Formicidae Pteromalidae Torymidae Trichogrammatidae

Aphytis spp., Aphelinus spp. Apis spp. Andricus kollari Hartig spp. Solenopsis invicta Buren, S. richteri Forel Acanthomyops spp. Catolaccus grandis Burks Torymus sinensis Kamijo, T. beneficus Yasumatsu and Kamijo Trichogramma minutum Riley spp.

Culicidae Culicidae Culicidae Culicidae Culicidae Culicidae Culicidae

Hymenoptera

Lab.

Field

x

x x x x

x x x x x x x x x x

x

x

x

x x x x

x x x

Publication (Michailova, 1998) (Polukonova and Beljanina, 2002) (Besansky et al., 1997) (Cornel et al., 2003) (Miao et al., 1988) (Taylor and Craig, 1985) (Taylor, 1990) (Thelwell et al., 2000) (Sherron and Rai, 1984) (Coyne and Orr, 1997) (Craig et al., 1997) (Dos Santos et al., 2001) (Pike et al., 2003) (Selivon et al., 1999) (Rao and DeBach, 1969) (Clarke et al., 2002) (Walker et al., 2002) (Shoemaker et al., 1996) (Umphrey and Danzmann, 1998) (Morales-Ramos et al., 2000) (Yara et al., 2000) (Pinto et al., 2003)


Order

81

82

K.R. Hopper et al.

Methods for prediction Field data on climatic distributions of candidates for introduction and native species at risk should allow assessment of whether they will become sympatric after introduction, assuming no barriers to dispersal. Computer programs, such as BIOCLIM (Nix, 1986) and GARP (Chen and Peterson, 2000), have been developed for climate matching and these may be useful for predicting distributions of biological control agents after introduction. If climatic distributions overlap but there are geographic or habitat barriers between the region of introduction and the region harbouring a native species at risk, the dispersal capacity of the candidate for introduction must be assessed. Given that climatic tolerances of introduced species may evolve, it would be useful to measure genetic variation in climatic tolerances in the material to be introduced. Methods for detection Field data on actual distribution of a biological control agent after introduction and spread will reveal which closely related native species are actually at risk of interbreeding. Case history Pak and Oatman (1982) and Glenn et al. (1997) measured development times for several populations of Trichogramma across California and Australia, respectively, and showed that species pairs or complexes were differently adapted to temperature regimes or to temperature requirements of their hosts.

Spatial and temporal barriers to mating Sympatric species that mate in different seasons or in different habitats will rarely, if ever, interbreed because they rarely meet (Feder et al., 1994; Bush and Smith, 1998; Tilmon et al., 1998). Temporal isolation could extend to differences in diurnal rhythm of courtship (Wood and Guttman,

1982; Feder et al., 1993; Morrow et al., 2000), and habitat specificity can involve differences in choice of host plant species for mating (Feder and Bush, 1989; Wood et al., 1999). Some courtship and mating may happen even with very little spatial or temporal overlap (Deverno et al., 1998; Haegele, 1999). Although rare courtship and mating are unlikely to affect demography, they may lead to introgression. Whether rare mating leads to introgression depends on post-zygotic barriers (inviability, sterility and reduced fitness of hybrids) and on the selective advantage or disadvantage conferred by introgressed sequences. New introgression is unlikely for augmentation of native species, unless the augmentation is in seasons or habitats where the augmented species does not ordinarily occur. Methods for prediction Field data on habitat and seasonal distributions of candidates for introduction and native species at risk will allow assessment of whether they will encounter one another in sympatry. For augmentative releases, the habitat and seasonal distribution of augmentation programmes can be used to predict which native species are at risk of interbreeding. Field and laboratory data on diurnal periodicity and on host plant fidelity of mating behaviour could show whether individuals of different species will encounter one another at smaller spatial and temporal scales. Methods for detection Field data on habitat and seasonal distributions of introduced agents and native species will show whether they are likely to encounter one another in sympatry. Because introgression may occur even with rare matings, survey for such introgression may prove useful. Techniques for detecting introgression are discussed below. Case history To assay seasonal activity of several species of Trichogramma over two years,


Thorpe (1984) placed Heliothis virescens Fabricius (Lepidoptera: Noctuidae) eggs in experimental plots of soybean, weedy vegetation and trees. Parasitism was temporally partitioned among four Trichogramma species in the first year, but was more evenly distributed in the second year.

83

court and couple only with conspecifics in the laboratory, this is likely to hold in the field as well. However, if candidate and native species court or couple with one another, this may or may not mean they will do so in the field. Methods for detection

Mate recognition Species whose mating periods and habitats more or less overlap must still recognize one another as potential mates. Mate recognition cues include colours, shapes, scents, songs and dances (Cibrian and Mitchell, 1991; Heady and Denno, 1991; Monti et al., 1995; Haegele, 1999; Dos Santos et al., 2001; Deering and Scriber, 2002). To what extent such cues are recognized between species varies with taxonomic group, phylogenetic proximity and whether species are sympatric or allopatric (Coyne and Orr, 1997). Partial recognition of cues which does not lead to mating or hybridization can still mean time wasted courting. Methods for prediction Field and laboratory data on mate recognition cues will reveal whether species are likely to recognize one another as mates. Laboratory mating trials will show whether they do indeed recognize one another as mates, how frequently they do so and to what degree (courtship could reach various levels of completion up to copulation). Candidates for introduction must be reared for at least one generation in the laboratory for host-range testing, identification, and clearing of hyperparasitoids and pathogens. Thus, introduction candidates include only species that will mate in the laboratory. Candidates for augmentation also must mate in the laboratory, otherwise they cannot be augmented. Given that native species at risk are closely related to candidates for introduction or augmentation, it is likely that they will mate in the laboratory as well. In any case, within-species crosses can serve as controls for between-species crosses. If candidate and native species

Field observations of courtship and mating in areas where introduced or augmented species and native species commonly cooccur would show whether they recognize one another as mates. Such observations are difficult with small, possibly nocturnal insects, especially if one or the other species is rare. Traps baited with virgin females (Davis et al., 1987; Brodeur and McNeil, 1994) may help in showing whether scents or songs are recognized between species. If interspecific courtship or mating is common enough to affect population dynamics, the effects may be measured by comparing reproductive rates or stage distributions before versus after introduction or augmentation, or in areas with versus without the introduced or augmented species. Case history Production of female offspring in matings between T. minutum and Trichogramma platneri Nagarkatti from different geographic regions is rare (Pinto et al., 2003). Stouthamer et al. (2000) measured mate choice between T. minutum and T. platneri in the laboratory by observing female behaviour in the presence of conspecific and heterospecific males. Although no female offspring resulted from mating with heterospecific males, females did not prefer males of either species. Thus releasing one species in the range of the other might affect population dynamics (Pinto et al., 2003).

Copulation and sperm use Species may recognize one another as mates sufficiently to attempt copulation,

84

K.R. Hopper et al.

but be unable to copulate normally because of morphological differences in genitalia (Arnqvist, 1998). Such differences can lead to injury or death of the female, or death of both partners if they become locked in copula (Sota and Kubota, 1998). Even if the partners survive, females may act as if they have been mated, even though no sperm has been transferred. If sperm is transferred, it may not fertilize heterospecific eggs (Jamart et al., 1995) or may be less competitive in multiple-mated females (Robinson et al., 1994; Howard, 1999). Females of some insect species cease to court, or reject males, after copulation or insemination (Allen et al., 1994; Fleischmann et al., 2001; Jang, 2002). Thus, if inseminated with heterospecific sperm first, these females may have reduced or no receptivity towards conspecific males. Interspecific mating may have no measurable effects, especially if females readily remate (Albuquerque et al., 1996). However, if interspecific matings are common, they could reduce the net reproductive rates of one or both species. The worst-case scenario would be where an introduced species, ineffective at controlling an abundant pest, was maintained at high numbers and mated with a rare native species whose females did not remate. This could lead to the native species being swamped with interspecific matings. Augmentative releases could have this effect where density was increased, but the effect should not extend beyond the dispersal range of the augmented species. A second scenario would be where an introduced species, while still rare, mated with a common native species, so that the introduced species was swamped with interspecific matings. At low densities, Allee effects, for example from failure to find appropriate mates, might become important and extinction possible (Hopper and Roush, 1993). In the second scenario, this would mean effort wasted in a failed introduction; in the first scenario, this would mean the loss of a native species. Such interactions are like pest control using sterile males, which has been effective in several cases (Gould and Schliekelman, 2004).

Methods for prediction Examination of genitalia might reveal whether morphological incompatibilities would be likely between candidates for introduction or augmentation and native species (but see Porter and Shapiro, 1990; Goulson, 1993; Eberhard, 2001). Laboratory crosses would show whether interspecific mating compromised intraspecific receptiveness and whether sperm was transferred and used interspecifically. Methods for detection To determine whether native or introduced females were sterilized without sperm transfer, one could collect females in the field, test whether they are receptive to mating with conspecific males in the laboratory, and then dissect the females to determine whether they carry sperm. To determine whether native or introduced females were sterilized by copulations with sperm transfer, one could collect females in the field, allow them to oviposit in the laboratory, and then dissect them to determine whether they carried sperm. Eggs which produced no progeny (or only male progeny for haplodiploids) would reveal that females had been sterilized. One would need to test females mated with conspecific males (either in the laboratory or in regions without the other species) to control for levels of sterility/infertility within species. As with courtship or mating without copulation or sperm transfer, one could measure demographic impacts by comparing population dynamics before versus after introduction or augmentation or in areas with versus without introduced or augmented species. Case history Genital morphology, correlated with reproductive incompatibility between Trichogramma species (Rohi and Pintureau, 2003), was used to estimate interspecific divergences among species groups (Pintureau, 1993). On the other hand, Nagarkatti and Fazaluddin (1973) found production of hybrids in the laboratory did not correlate with morphological, and in


particular genitalic, similarity, geographic proximity or habitat similarity. Although heterogamic insemination was frequent, production of interspecific hybrids was rare, and most cases of hybrid progeny were unidirectional. Stouthamer et al. (2000) describe methods for observing sperm transfer, and Damiens et al. (2002) describe a method for measuring viability of sperm in spermathecae of females.

Impacts of Interbreeding Hybrid progeny: inviability and sterility If species mate and reproduce, the hybrid progeny may show reduced viability or fertility. If interspecific matings are common, production of inviable progeny could affect population dynamics like copulation, which effectively sterilizes females. Thus, the two scenarios described under the section on copulation apply here as well: (1) a rare native species being swamped by hybridization with an ineffective introduced biological control agent or with an augmented native species, or (2) a newly introduced species being swamped by hybridization with a common native species. If hybrid progeny are viable but sterile, these two scenarios would be worsened because hybrids would also mate with, and thus effectively sterilize, individuals of one or both species. The most abundant species would mostly mate intraspecifically, but the least abundant species would either mate interspecifically or with hybrids. This interaction would be like the project to control Heliothis virescens using sterile-male hybrids from crosses between H. virescens and H. subflexa Guenée (King et al., 1985; Proshold et al., 1986). Although this project had mixed success, it showed sufficient promise at suppression to be pursued for a decade, suggesting that interspecies hybridization might reduce abundances of either native species or introduced biological control agents. The Heliothis-hybrid project produced a series of mathematical models (Roush and Schneider, 1985; Laster et al., 1996) that could be used to analyse

85

the likely effects of sterile hybrids between introduced or augmented species and native species. Methods for prediction Laboratory crosses could show whether hybrid progeny are produced and whether these progeny are inviable or sterile. However, fertility and survival may be intermediate, with either all hybrids showing intermediate fertility or survival or with some crosses producing inviable/sterile hybrids and others producing fit hybrids (Oliver, 1979; Collins, 1997; Coyne and Orr, 1997; Presgraves, 2002). Withinspecies crosses would provide controls for level and between-family distribution of survival or fertility. Methods for detection To determine whether native or introduced females are producing inviable hybrid progeny, one could collect females from the field, allow them to oviposit in the laboratory, dissect the females to determine whether mated, and then measure the number of progeny (or female progeny for haplodiploids) reaching adulthood. One would need controls of females known to be mated with conspecific males. These could be obtained from areas where the species did not overlap or from laboratory crosses. To determine whether native or introduced females are producing viable hybrids, one could search for hybrids in field collections. Hybrids would be easiest to find where introduced and native species are about equal in abundance. Hybrids could be identified either by phenotype using, for instance, the hybrid character index (Anderson, 1936; Howard et al., 1993) or genotype (Nason and Ellstrand, 1993; Anderson and Thompson, 2002). If hybrids show readily identifiable morphological phenotypes, clearly distinguishable from both introduced and native parents, screening by phenotype might be simplest. On the other hand, hybrids between closely related species might be difficult to identify by

86

K.R. Hopper et al.

phenotype. In this case, hybrids could be detected as heterozygotes of fixed molecular differences between native species and introduced or augmented species. Once fixed molecular differences between native and introduced or augmented species are established, the presence of such heterozygotes could easily be detected. Insertions/deletions in nuclear ribosomal genes like ITS1 and ITS2, or in introns flanked by conserved exons, might be easiest to detect because they would not require sequencing (e.g. Zhu et al., 2000). Single nucleotide polymorphisms (SNPs) that distinguish native versus introduced or augmented haplotypes could also be detected without sequencing (e.g. Morlais and Severson, 2002). Rare hybrids would be difficult to detect by either phenotypic or genotypic screening. However, for hybridization without introgression to affect population dynamics of native or introduced species, hybrids would have to be common. One could measure effects of introduced or augmented species on dynamics of native species by comparing populations before versus after introduction or augmentation, or in areas with versus without the introduced or augmented species. Measuring effects of native species on demography of introduced species would be more difficult, unless there were areas in the region of introduction where hybridization was absent or at least less common. In the latter case, one could compare introduced species dynamics with and without hybridization with the native species. Case history Although interspecies hybrids were rare, Nagarkatti and Fazaluddin (1973) found in all instances where hybrids were produced, hybrids were viable and fertile.

so even if hybrids are sterile. The most worrisome shift in host use would be to attack species used by the native species yielding the hybrid. Methods for prediction If the hosts of the native species producing hybrids are of concern, the host range of hybrids could be measured in laboratory experiments like those for evaluation of host use by candidates for introduction or augmentation (see van Lenteren et al., Chapter 3, this volume). Methods for detection Collecting from hosts of native species at risk for interbreeding should reveal whether hybrids are parasitizing these hosts. The techniques for detecting such hybrids are discussed above under ‘Hybrid progeny: inviability and sterility’.

Hybrid speciation Hybrids between species may not cross with parental species but cross among themselves, which could give rise to a new species (Arnold, 1997; Barton, 2001). Such hybrid speciation is much more likely in plants than in insects (Rieseberg et al., 1995; Rieseberg, 1997), but is being discovered in a growing number of animals (Bullini, 1994). Augmentation of native species is unlikely to produce new, hybrid species. If a hybrid species were to result from a biological control introduction, it would be as if two species had been introduced, one with known traits and the other with some combination of traits from the introduced and native parents. Methods for prediction

Hybrid progeny: host range shifts If viable hybrids are produced, they could affect unexpected non-target species if hybrid host range differed from that of the introduced or augmented species. This is

Laboratory crosses among hybrids and between hybrids and their parental species could show whether hybrids would be most likely to cross among themselves or backcross to the parental species after introduction.


Methods for detection To determine whether hybrids were breeding among themselves or backcrossing to the introduced or native parents, one could collect insects from the field and screen for hybrid phenotypes and genotypes. Hybrids mating among themselves would show segregation of genes from both parents. Segregation should be detectable with phenotypic and molecular markers, even if some hybrid genotypes are more favoured than others.

Reproductive character displacement If species commonly hybridize, but hybrids have lower fitness than either parental species, reproductive traits of one or both species may diverge (Dobzhansky, 1940). Such reproductive character displacement has often been invoked in discussions of sympatric speciation and reinforcement after secondary contact between allopatric species (McLain, 1986; Bordenstein et al., 2000; Kawano, 2002). Some are sceptical about the likelihood of reinforcement (Moore, 1957; Mayr, 1963; Barton and Hewitt, 1981), but recent models and data of sympatric speciation (Via, 2001) and of reinforcement (Howard, 1993) suggest that reproductive character displacement may be more common than many have thought. Post-introduction evolution in reproductive traits of introduced species is probably not of concern, unless it would affect success in biological control, which seems unlikely. On the other hand, some would consider evolution in reproductive traits of native species undesirable (Simberloff and Stiling, 1996; Mooney and Cleland, 2001), although such evolution would not necessarily mean changes in abundance.

Introgression Fertile hybrids between species may backcross to either parental species and thus introgress DNA sequences from one species into another (Anderson, 1953; Dowling and

87

Secor, 1997). This is true even if hybrid fitness is low and successful backcrosses rare (Barton, 2001). However, the genomes of species sufficiently close to hybridize are quite similar (Hewitt, 1988; Barton, 2001), so that many introgressed sequences will have no effect, either not changing sequences or not changing function, and thus not changing fitness. Augmentation of native species is unlikely to increase introgression much. However, native and introduced species presumably differ, at least in host range or impact on the target pest; otherwise the candidate for introduction would not be under consideration. The fate and impact of introgressed sequences depends on the selective advantage or disadvantage they confer, the frequency of introgression, and dispersal rates (Barton and Gale, 1993; Barton, 2001). If hybrids and backcrosses are common, neutral and even mildly deleterious genes could become common in the area of contact, although they would be unlikely to spread far beyond the hybrid zone (Barton and Hewitt, 1981). As discussed above under ‘Hybrid progeny: inviability and sterility’, which species would be most affected depends on relative abundances. A rare native, swamped by backcrosses with hybrids from a common introduced species, could have high levels of introgression, and the same applies to a rare introduced species swamped by backcrosses with hybrids from a common native. High levels of introgression would be relatively easy to detect using molecular markers. Introgressed sequences that are strongly deleterious, either through direct effects on traits fitness components or through breakup of co-adapted gene complexes, would be strongly selected against and thus unlikely to persist or spread. Thus, the major effect of introgression of deleterious sequences would be demographic. If hybrids and successful backcrosses are rare, introgressed sequences would be unlikely to persist or spread unless they are selectively advantageous (Barton and Hewitt, 1985; Barton and Gale, 1993; Linder et al., 1998). However, if an introgressed sequence is selectively advantageous, it

88

K.R. Hopper et al.

could rapidly sweep to fixation with modest levels of selective advantage and dispersal (Barton, 2001). Unfortunately, such a selective sweep could be very difficult to detect if it involved a small, unknown sequence affecting a trait not previously measured. On the other hand, introgression of sequences affecting previously measured traits, like host specificity or climatic tolerances, would be relatively easy to detect. Introgression of sequences affecting traits like host use or climatic tolerances could have major, and perhaps unwanted, consequences. Introgression of sequences affecting climatic tolerances or diapause conditions could allow range expansions of native species or increase the realized ranges of an introduced species. On the other hand, introgressed genes affecting climatic tolerances could act like conditional lethals, spreading because of fitness advantage under current conditions and then causing heavy mortality when conditions change. Introgression of such genes resembles proposals for genetic control, which although seldom implemented, show much promise for pest management (Gould and Schliekelman, 2004). Nonetheless, such a catastrophic outcome seems unlikely by chance given the similarity between genomes of species that will hybridize. Introgression of sequences affecting host range raises the most worrisome and plausible scenarios for interbreeding. Sequences introgressed from an introduced species into a native species could allow the native species to attack species beyond its original range, including the target pest. The latter would not be bad in itself, and indeed might provide useful control, but some hold that any such introgression-driven evolution is a form of environmental pollution and thus should be avoided (Mooney and Cleland, 2001; Allendorf and Lundquist, 2003). Sequences introgressed from a native species into an introduced species could cause a rapid shift in host range, allowing the introduced species to attack hosts of the native species. Given that host specificity is essential for the safety of biological control

introductions, introgression-driven evolution of host range, especially to attack native species, is clearly undesirable. The likelihood of such introgressive changes in host range is unknown; no examples are available in the literature. Methods for prediction Laboratory crosses could show with what frequency backcross progeny are produced from crosses of hybrids with either candidates for introduction or native species. If backcross progeny are produced, one could measure their host range, climatic tolerances, mating behaviour and other traits of interest. Within-species crosses would be needed as controls for expected levels of traits. Methods for detection If backcrosses are common, one could measure introgression using molecular markers. If backcrosses are rare, it will be difficult to measure introgression using molecular markers, unless one has markers for specific genes of interest. For rare introgression, measurement of phenotypic changes will be easier. Beside changes in host range and climatic tolerances measured in the laboratory crosses, one could measure morphological traits after various generations of backcrossing to determine whether introgression could be detected by screening morphological phenotypes. One could measure demographic effects of introgression from introduced species into native species by comparing population dynamics before versus after introduction, or in populations where introgression had or had not occurred. The latter approach depends on being able to detect introgression. Measuring demographic effects of introgression from native species into introduced species would be more difficult, unless there were areas in the region of introduction where introgression was absent. In this case, one could compare introduced species dynamics with and without introgression with the native species.


89

Case history

Recommendations and Conclusions

Because T. minutum and T. platneri hybridized occasionally in the laboratory and there was concern about gene flow between these species in the wild, Pinto et al. (2003) tested for introgression where these species are sympatric in the Pacific North-west and found no introgression of species-specific alleles at the Pgm locus. Although hybrids are expected to differ phenotypically from parents and thus be detectable, Nagarkatti and Fazaluddin (1973) found hybrids invariably resembled the maternal parent. As evidence that introgression could be deleterious, hybridizing geographical populations of Trichogramma and selecting hybrids for tolerance of temperature extremes produced a weak response and actually reduced parasitism in the laboratory (Ashley et al., 1974).

We organized the tests described above into flowcharts for predicting the risks of interbreeding from introduction (Fig. 5.1) and augmentation (Fig. 5.2), and for assessing impacts of interbreeding with introduced (Fig. 5.3) or augmented species (Fig. 5.4). Decision makers must realize that these flowcharts are extremely schematic. The details of the biology of each biological control candidate, and what is known about that biology, may require modifications in the flowcharts or in the tests proposed above. The major differences between procedures for introductions (Figs 5.1 and 5.3) and augmentation (Figs 5.2 and 5.4) are that those for introductions address the risk of introgression of novel genes, while those for augmentation concentrate on the demographic effects of mating and hybridization. The

Closely related (<same genus)?

No

Yes Likely geographical overlap?

No

Yes Likely habitat/seasonal overlap?

No

Yes Recognize as mates?

No

Yes Hybridize?

No

Yes Hybridize often (>1/10)? No

Yes

Hybrids viable and fertile? Yes

Further study No Accept

Further study

Fig. 5.1. Pre-introduction tests to predict interbreeding between species introduced for biological control and native species. See text for description of tests.

90

K.R. Hopper et al.


No

Yes Likely geographical overlap?

No

Yes Likely habitat/seasonal overlap?

No

Yes Recognize as mates?

No

Yes Hybridize?

No

Yes Hybridize often (>1/2)? Yes

No Accept

Further study

Fig. 5.2. Tests to predict interbreeding between native species augmented for biological control and other native species. See text for description of tests.

major difference between predicting impacts (Figs 5.1 and 5.2) and assessing impacts (Figs 5.3 and 5.4) is that the latter involve field measurements of mating, hybridization and introgression. Where post-introduction tests show effects on populations of nontarget species, further releases of introduced species and augmentation of native species should be stopped, and similar candidates should be avoided in the future. To illustrate how one might proceed with these flowcharts, we will use our results on the Aphelinus varipes complex (Hymenoptera: Aphelinidae) (K.R. Hopper, J.B. Woolley, J.M. Heraty, A.M.I. Farias, S.C. Britch, unpublished results). The A. varipes complex comprises a group of sibling species in Eurasia. One species from the Republic of Georgia parasitizes Diuraphis noxia (Mordvilko) (Hemiptera: Aphididae), the Russian wheat aphid. If D. noxia were accidentally introduced into Japan and became a pest (as has occurred in the United States), this Georgian species would be a candidate for introduction into

Japan. However, a species in the A. varipes complex, which does not parasitize D. noxia, already occurs in Japan. Because the Georgian and Japanese species are closely related, the answer to the first question in Fig. 5.1 would be ‘Yes’. The climate in Georgia where the parasitoids were collected matches fairly well the target climate in Japan (Walter and Lieth, 1967), so the species are likely to overlap in geographical range after introduction, and the answer to the second question in Fig. 5.1 would be ‘Yes’. Because they overlap broadly in host range, with the exception of D. noxia, they are also likely to occur in the same habitats, and the answer to the third question in Fig. 5.1 would be ‘Yes’. Their DNA sequences differ across several genes, indicating they have had separate evolutionary histories for several hundred thousand years, but these species readily mate and produce viable offspring in laboratory experiments. Thus, the answers to the remaining questions in Fig. 5.1 would also be ‘Yes’. Given that the introgression



91

No

Yes Geographical overlap?

No

Yes Habitat/seasonal overlap?

No

Yes Recognize as mates in field? Yes

No

Effects on populations?

Hybridize in field?

No No

Yes No

Hybridize often (>1/10)? No

Yes


Hybrids viable and fertile?

No No

Yes Introgression occurs? Yes

No


No No impact detected

Fig. 5.3. Post-introduction tests to measure occurrence and impact of interbreeding between species introduced for biological control and native species. See text for description of tests. Where there are effects on populations of non-target species, further releases should be stopped and similar candidates should be avoided in the future.

seems likely, further study would be needed according to Fig. 5.1. However, because the species differ in host use and introgression could lead to an evolutionary shift in host use, in either the introduced or native species, we would recommend against releasing the Georgian species in Japan, particularly because there are other candidates with narrower host ranges that do not mate with the Japanese species. In our opinion, the risks are small of large impacts from interbreeding between native species and insects used in biological control. But data are lacking about both the likelihood and impact of interbreeding, so more research is needed. Fortunately,

this research will not only improve the safety of biological control, but will also shed light on the behaviour, ecology and genetics of courtship, mating and hybridization, and thus on the mechanisms of speciation.

Acknowledgements We thank the participants of the Engelberg workshop of June 2004 for providing insightful discussions, and an anonymous reviewer for comments on the manuscript, and USDA-ARS and INRA for their support during the preparation of this chapter.

92

K.R. Hopper et al.


No

Yes Geographical overlap?

No

Yes Habitat/seasonal overlap?

No

Yes Recognize as mates in field? Yes


Hybridize in field?

No No No

Yes Hybridize often (>1/10)?

No

Yes

No


No impact detected

Fig. 5.4. Tests to measure occurrence and impact of interbreeding between species augmented for biological control and native species. See text for description of tests. Where there are effects on populations of non-target species, augmentation should be stopped and further studies conducted.

References Albuquerque, G.S., Tauber, C.A. and Tauber, M.J. (1996) Postmating reproductive isolation between Chrysopa quadripunctata and Chrysopa slossonae: Mechanisms and geographic variation. Evolution 50, 1598–1606. Allen, G.R., Kazmer, D.J. and Luck, R.F. (1994) Post copulatory male behaviour, sperm precedence and multiple mating in a solitary parasitoid wasp. Animal Behaviour 48, 635–644. Allendorf, F.W. and Lundquist, L.L. (2003) Introduction: Population biology, evolution, and control of invasive species. Conservation Biology 17, 24–30. Anderson, E. (1936) Hybridization in American Tradescantias. Annals of the Missouri Botanical Gardens 23, 511–525. Anderson, E. (1953) Introgressive hybridization. Biological Reviews 28, 280–307. Anderson, E.C. and Thompson, E.A. (2002) A model-based method for identifying species hybrids using multilocus genetic data. Genetics 160, 1217–1229. Arnold, M.L. (1997) Natural Hybridization and Evolution. Oxford University Press, New York. Arnqvist, G. (1998) Comparative evidence for the evolution of genitalia by sexual selection. Nature 393, 784–785. Ashley, T.R., Gonzalez, D. and Leigh, T.F. (1974) Selection and hybridization of Trichogramma. Environmental Entomology 3, 43–48. Barton, N.H. (2001) The role of hybridization in evolution. Molecular Ecology 10, 551–568. Barton, N.H. and Gale, K.S. (1993) Genetic analysis of hybrid zones. In: Harrison, R.G. (ed.) Hybrid Zones and the Evolutionary Process. Oxford University Press, New York. Barton, N.H. and Hewitt, G.M. (1981) Hybrid zones and speciation. In: Atchley, W.R. and Woodruff, D.S. (eds) Evolution and Speciation: Essays in Honor of M. J. D. White. Cambridge University Press, Cambridge, UK. Barton, N.H. and Hewitt, G.M. (1985) Analysis of hybrid zones. Annual Review of Ecology and Systematics 16, 113–148.


93

Besansky, N.J., Lehmann, T., Fahey, G.T., Fontenille, D., Braack, L., Hawley, W.A. and Collins, F.H. (1997) Patterns of mitochondrial variation within and between African malaria vectors, Anopheles gambiae and Anopheles arabiensis, suggest extensive gene flow. Genetics 147, 1817–1828. Bordenstein, S.R., Drapeau, M.D. and Werren, J.H. (2000) Intraspecific variation in sexual isolation in the jewel wasp Nasonia. Evolution 54, 567–573. Brennan, J.M. and Fairbairn, D.J. (1995) Clinal variation in morphology among eastern populations of the waterstrider, Aquarius remigis Say (Hemiptera: Gerridae). Biological Journal of the Linnean Society 54, 151–171. Brodeur, J. and McNeil, J.N. (1994) Seasonal ecology of Aphidius nigripes (Hymenoptera: Aphidiidae), a parasitoid of Macrosiphum euphorbia (Homoptera: Aphididae). Environmental Entomology 23, 292–298. Bullini, L. (1994) Origin and evolution of animal hybrid species. Trends in Ecology and Evolution 9, 422–426. Bush, G.L. and Smith, J.J. (1998) The genetics and ecology of sympatric speciation: A case study. Researches on Population Ecology 40, 175–187. Chen, G.J. and Peterson, A.T. (2000) A new technique for predicting distribution of terrestrial vertebrates using inferential modeling. Zoological Research 21, 231–237. Cibrian, T.J. and Mitchell, E.R. (1991) Courtship behaviour of Heliothis subflexa (Gn.) (Lepidoptera: Noctuidae) and associated backcross insects obtained from hybridization with Heliothis virescens (F.). Environmental Entomology 20, 419–426. Clarke, K.E., Rinderer, T.E., Franck, P., Quezada-Euan, J.J.G. and Oldroyd, B.P. (2002) The Africanization of honeybees (Apis mellifera L.) of the Yucatan: A study of a massive hybridization event across time. Evolution 56, 1462–1474. Clausen, C.P. (1978) Introduced Parasites and Predators of Arthropod Pests and Weeds: a World Review. Agriculture Handbook No. 480. Agricultural Research Service, USDA, Washington, DC. Collins, M.M. (1997) Hybridization and speciation in Hyalophora (Insecta: Lepidoptera: Saturniidae): A reappraisal of W. R. Sweadner’s classic study of a hybrid zone. Annals of Carnegie Museum No. 14. 66, 411–455. Cornel, A.J., McAbee, R.D., Rasgon, J., Stanich, M.A., Scott, T.W. and Coetzee, M. (2003) Differences in extent of genetic introgression between sympatric Culex pipiens and Culex quinquefasciatus (Diptera: Culicidae) in California and South Africa. Journal of Medical Entomology 40, 36–51. Cox, G.W. (2004) Alien Species and Evolution. Island Press, Washington, DC. Coyne, J.A. and Orr, H.A. (1997) ‘Patterns of speciation in Drosophila’ revisited. Evolution 51, 295–303. Craig, T.P., Horner, J.D. and Itami, J.K. (1997) Hybridization studies on the host races of Eurosta solidaginis: Implications for sympatric speciation. Evolution 51, 1552–1560. Damiens, D., Bressac, C., Brillard, J.P. and Chevrier, C. (2002) Qualitative aspects of sperm stock in males and females from Eupelmus orientalis and Dinarmus basalis (Hymenoptera: Chalcidoidea) as revealed by dual fluorescence. Physiological Entomology 27, 97–102. Danforth, B.N., Mitchell, P.L. and Packer, L. (1998) Mitochondrial DNA differentiation between two cryptic Halictus (Hymenoptera: Halictidae) species. Annals of the Entomological Society of America 91, 387–391. Davis, F.M., Bird, T.G., Sloderbeck, P.E., Lewis, B.E., Yochim, R.S., Knutson, A.E., Ng, S.S., Gallardo, J.L., Pedroza, A.S. and Mihm, J.A. (1987) Southwestern corn borer attractiveness to synthetic pheromone. Southwestern Entomologist 12, 57–65. Deering, M.D. and Scriber, J.M. (2002) Field bioassays show heterospecific mating preference asymmetry between hybridizing North American Papilio butterfly species (Lepidoptera: Papillonidae). Journal of Ethology 20, 25–33. Deverno, L.L., Smith, G.A. and Harrison, K.J. (1998) Randomly amplified polymorphic DNA evidence of introgression in two closely related sympatric species of Coniferophagous choristoneura (Lepidoptera: Tortricidae) in Atlantic Canada. Annals of the Entomological Society of America 91, 248–259. Dobzhansky, T. (1940) Speciation as a stage in evolutionary divergence. American Naturalist 74, 312–321. Dobzhansky, T. (1970) Genetics of the Evolutionary Process. Columbia, New York.

94

K.R. Hopper et al.

Dos Santos, P., Uramoto, K. and Matioli, S.R. (2001) Experimental hybridization among Anastrepha species (Diptera: Tephritidae): Production and morphological characterization of F1 hybrids. Annals of the Entomological Society of America 94, 717–725. Dowling, T.E. and Secor, C.L. (1997) The role of hybridization and introgression in the diversification of animals. Annual Review of Ecology and Systematics 28, 593–619. Eberhard, W.G. (2001) Species-specific genitalic copulatory courtship in sepsid flies (Diptera: Sepsidae: Microsepsis) and theories of genitalic evolution. Evolution 55, 93–102. Endler, J.A. (1977) Geographic Variation, Speciation, and Clines. Princeton University Press, Princeton, New Jersey. Ewel, J.J., O’Dowd, D.J., Bergelson, J., Daehler, C.C., D’Antonio, C.M., Gomez, L.D., Gordon, D.R., Hobbs, R.J., Holt, A., Hopper, K.R., Hughes, C.E., Lahart, M., Leakey, R.R.B., Lee, W.G., Loope, L.L., Lorence, D.H., Louda, S.M., Lugo, A.E., McEvoy, P.B., Richardson, D.M. and Vitousek, P.M. (1999) Deliberate introductions of species: Research needs–benefits can be reaped, but risks are high. Bioscience 49, 619–630. Feder, J.L. and Bush, G.L. (1989) Gene frequency clines for host races of Rhagoletis pomonella in the midwestern USA. Heredity 63, 245–266. Feder, J.L., Hunt, T.A. and Bush, G.L. (1993) The effects of climate, host-plant phenology and host fidelity on the genetics of apple and hawthorn infesting races of Rhagoletis pomonella. Entomologia Experimentalis et Applicata 69, 117–135. Feder, J.L., Opp, S.B., Wlazlo, B., Reynolds, K., Go, W. and Spisak, S. (1994) Host fidelity is an effective premating barrier between sympatric races of the apple maggot fly. Proceedings of the National Academy of Sciences of the USA 91, 7990–7994. Fleischmann, I., Cotton, B., Choffat, Y., Spengler, M. and Kubli, E. (2001) Mushroom bodies and postmating behaviors of Drosophila melanogaster females. Journal of Neurogenetics 15, 117–144. Frank, J.H. and McCoy, E.D. (1995) Introduction to insect behavioral ecology – the good, the bad, and the beautiful – non-indigenous species in Florida. Florida Entomologist 78, 1–15. Glenn, D.C., Hercus, M.J. and Hoffmann, A.A. (1997) Characterizing Trichogramma (Hymenoptera: Trichogrammatidae) species for biocontrol of light brown apple moth (Lepidoptera: Tortricidae) in grapevines in Australia. Annals of the Entomological Society of America 90, 128–137. Gould, F. and Schliekelman, P. (2004) Population genetics of autocidal control and strain replacement. Annual Review of Entomology 49, 193–217. Goulson, D. (1993) Variation in the genitalia of the butterfly Maniola jurtina (Lepidoptera: Satyrinae). Zoological Journal of the Linnean Society 107, 65–71. Haegele, K. (1999) Hybrid syndrome-induced postzygotic reproductive isolation: A second reproduction barrier in Chironomus thummi (Diptera: Chironomidae). Journal of Zoological Systematics and Evolutionary Research 37, 161–164. Harrison, R.G. (1986) Pattern and process in a narrow hybrid zone. Heredity 56, 337–350. Heady, S.E. and Denno, R.F. (1991) Reproductive isolation in Prokelisia planthoppers (Homoptera: Delphacidae): Acoustic differentiation and hybridization failure. Journal of Insect Behavior 4, 367–390. Hewitt, G.M. (1988) Hybrid zones – natural laboratories for evolutionary studies. Trends in Ecology and Evolution 3, 158–167. Hopper, K.R. and Roush, R.T. (1993) Mate finding, dispersal, number released, and the success of biological-control introductions. Ecological Entomology 18, 321–331. Howard, D.J. (1986) A zone of overlap and hybridization between two ground cricket species. Evolution 40, 34–43. Howard, D.J. (1993) Reinforcement: Origin, dynamics, and fate of an evolutionary hypothesis. In: Harrison, R.G. (ed.) Hybrid Zones and the Evolutionary Process. Oxford University Press, New York. Howard, D.J. (1999) Conspecific sperm and pollen precedence and speciation. Annual Review of Ecology and Systematics 30, 109–132. Howard, D.J., Waring, G.L., Tibbets, C.A. and Gregory, P.G. (1993) Survival of hybrids in a mosaic hybrid zone. Evolution 47, 789–800. Hoy, M.A., Jeyaprakash, A., Morakote, R., Lo, P.K.C. and Nguyen, R. (2000) Genomic analyses of two populations of Ageniaspis citricola (Hymenoptera: Encyrtidae) suggest that a cryptic species may exist. Biological Control 17, 1–10. Huxel, G.R. (1999) Rapid displacement of native species by invasive species: Effects of hybridization. Biological Conservation 89, 143–152.


95

Jamart, J.A., Casares, P., Carracedo, M.C. and Pineiro, R. (1995) Consequences of homospecific and heterospecific rapid remating on the fitness of Drosophila melanogaster females. Journal of Insect Physiology 41, 1019–1026. Jang, E.B. (2002) Physiology of mating behavior in Mediterranean fruit fly (Diptera: Tephritidae): Chemoreception and male accessory gland fluids in female post-mating behavior. Florida Entomologist 85, 89–93. Kawano, K. (2002) Character displacement in giant rhinoceros beetles. American Naturalist 159, 255–271. King, E.G., Hartley, G.G., Martin, D.F. and Laster, M.L. (1985) Large-scale rearing of a sterile backcross of the tobacco budworm Heliothis virescens (Lepidoptera: Noctuidae). Journal of Economic Entomology 78, 1166–1172. Kohlmann, B.C. and Shaw, D.D. (1991) The effect of a partial barrier on the movement of a hybrid zone. Evolution 45, 1606–1617. Laster, M.L., Hardee, D.D. and Schneider, J.C. (1996) Heliothis virescens (Lepidoptera: Noctuidae): Influence of sterile backcross releases on suppression. Southwestern Entomologist 21, 433–444. Linder, C.R., Taha, I., Seiler, G.J., Snow, A.A. and Rieseberg, L.H. (1998) Long-term introgression of crop genes into wild sunflower populations. Theoretical and Applied Genetics 96, 339–347. Mayr, E. (1963) Animal Species and Evolution. Belknap Press, Cambridge, Massachusetts. McLain, D.K. (1986) Niche differentiation and the evolution of ethological isolation in a soldier beetle hybrid zone. Oikos 47, 159–167. Miao, J., Pan, J. and Jang, W. (1988) Hybridization and chromosome observations on six species of the Anopheles hyrcanus group in China (Diptera: Culicidae). Zoological Research 9, 231–238. Michailova, P.V. (1998) Cytogenetic analysis of a hybrid, Glyptotendipes pallens (Mg.) ⫻ Glyptotendipes glaucus (Mg.) (Diptera: Chironomidae): Evolutionary considerations. Journal of Zoological Systematics and Evolutionary Research 36, 185–189. Monti, L., Lalanne-Cassou, B., Lucas, P., Malosse, C. and Silvain, J.F. (1995) Differences in sex pheromone communication systems of closely related species: Spodoptera latifascia (Walker) and S. descoinsi (Lalanne-Cassou and Silvain) (Lepidoptera: Noctuidae). Journal of Chemical Ecology 21, 641–660. Mooney, H.A. and Cleland, E.E. (2001) The evolutionary impact of invasive species. Proceedings of the National Academy of Sciences of the USA 98, 5446–5451. Moore, J.A. (1957) An embryologist’s view of the species concept. In: Mayr, E. (ed.) The Species Problem. American Association for the Advancement of Science, Washington, DC. Moore, W.S. (1977) An evaluation of narrow hybrid zones in vertebrates. Quarterly Review of Biology 52, 263–277. Morales-Ramos, J.A., Rojas, M.G. and King, E.G. (2000) Differences in reproductive potential of two populations of Catolaccus grandis (Hymenoptera: Pteromalidae) and their hybrids. Florida Entomologist 83, 137–145. Morlais, I. and Severson, D.W. (2002) Complete mitochondrial DNA sequence and amino acid analysis of the cytochrome-c oxidase subunit I (COI) from Aedes aegypti. DNA Sequence 13, 123–127. Morrow, J., Scott, L., Congdon, B., Yeates, D., Frommer, M. and Sved, J. (2000) Close genetic similarity between two sympatric species of tephritid fruit fly reproductively isolated by mating time. Evolution 54, 899–910. Nagarkatti, S. and Fazaluddin, M. (1973) Biosystematic studies on Trichogramma species (Hymenoptera: Trichogrammatidae). 2. Experimental hybridization between some Trichogramma spp. from the New World. Systematic Zoology 22, 103–117. Nason, J.D. and Ellstrand, N.C. (1993) Estimating the frequencies of genetically distinct classes of individuals in hybridized populations. Journal of Heredity 84, 1–12. Nix, H.A. (1986) A biogeographic analysis of the Australina elapid snakes. In: Longmore, R. (ed.) Atlas of Elapid Snakes, Australian Flora and Fauna Series No. 7, 4–15. Australian Government Publishing Service, Canberra. Nyman, T. (2002) The willow bud galler Euura mucronata Hartig (Hymenoptera: Tenthredinidae): One polyphage or many monophages? Heredity 88, 288–295. Oliver, C.G. (1979) Genetic differentiation and hybrid viability within and between some Lepidoptera species. American Naturalist 114, 681–694. Pak, G.A. and Oatman, E.R. (1982) Comparative life table, behavior and competition studies of Trichogramma brevicapillum and Trichogramma pretiosum. Entomologia Experimentalis et Applicata 32, 68–79.

96

K.R. Hopper et al.

Pike, N., Wang, W.Y.S. and Meats, A. (2003) The likely fate of hybrids of Bactrocera tryoni and Bactrocera neohumeralis. Heredity 90, 365–370. Pinto, J.D., Stouthamer, R., Platner, G.R. and Oatman, E.R. (1991) Variation in reproductive compatibility in Trichogramma and its taxonomic significance (Hymenoptera, Trichogrammatidae). Annals of the Entomological Society of America 84, 37–46. Pinto, J.D., Kazmer, D.J., Platner, G.R. and Sassaman, C.A. (1992) Taxonomy of the Trichogramma minutum complex (Hymenoptera, Trichogrammatidae) – allozymic variation and its relationship to reproductive and geographic data. Annals of the Entomological Society of America 85, 413–422. Pinto, J.D., Platner, G.R. and Stouthamer, R. (2003) The systematics of the Trichogramma minutum species complex (Hymenoptera: Trichogrammatidae), a group of important North American biological control agents: The evidence from reproductive compatibility and allozymes. Biological Control 27, 167–180. Pintureau, B. (1993) Morphometric analysis of the genus Trichogramma Westwood (Hymenoptera, Trichogrammatidae) in Europe. Canadian Entomologist 125, 367–378. Polukonova, N.V. and Beljanina, S.I. (2002) On the possibility of hybridogenesis in the origin of midge Chironomus usenicus Loginova et Beljanina (Chironomidae: Diptera). Genetika 38, 1635–1640. Porter, A.H. and Shapiro, A.M. (1990) Lock and key hypothesis: Lack of mechanical isolation in a butterfly (Lepidoptera: Pieridae) hybrid zone. Annals of the Entomological Society of America 83, 107–114. Presgraves, D.C. (2002) Patterns of postzygotic isolation in Lepidoptera. Evolution 56, 1168–1183. Proshold, F.I., King, E.G. and Hartley, G.G. (1986) Survival, emergence, and release of laboratoryreared backcross tobacco budworm Heliothis virescens (Lepidoptera: Noctuidae) pupae shipped from Stoneville, Mississippi to St. Croix, USA (Virgin Islands). Journal of Economic Entomology 79, 541–544. Rao, S.V. and DeBach, P. (1969) Experimental studies on hybridization and sexual isolation between some Aphytis species (Hymenoptera: Aphelinidae). III. The significance of reproductive isolation between interspecific hybrids and parental species. Evolution 23, 525–533. Rawlings, S.P. (1985) The genetics of hybrid sterility between subspecies of the complex of Glossina morsitans (Diptera: Glossinidae). Bulletin of Entomological Research 75, 689–700. Rieseberg, L.H. (1997) Hybrid origins of plant species. Annual Review of Ecology and Systematics 28, 359–389. Rieseberg, L.H., Vanfossen, C. and Desrochers, A.M. (1995) Hybrid speciation accompanied by genomic reorganization in wild sunflowers. Nature 375, 313–316. Robinson, T., Johnson, N.A. and Wade, M.J. (1994) Postcopulatory, prezygotic isolation: Intraspecific and interspecific sperm precedence in Tribolium spp., flour beetles. Heredity 73, 155–159. Rohi, L. and Pintureau, B. (2003) Are Trichogramma bourarachae and the perkinsi species group really distinct from Trichogramma buesi and the pintoi group, respectively? Journal of Applied Entomology 127, 265–268. Roush, R.T. and Schneider, J.C. (1985) An analytical model for genetic-control of Heliothis virescens incorporating the effects of sterile males. Theoretical and Applied Genetics 71, 472–477. Sagarra, L.A. and Peterkin, D.D. (1999) Invasion of the Caribbean by the hibiscus mealybug, Maconellicoccus hirsutus Green (Homoptera: Pseudococcidae). Phytoprotection 80, 103–113. Selivon, D., Perondini, A.L.P. and Morgante, J.S. (1999) Haldane’s rule and other aspects of reproductive isolation observed in the Anastrepha fraterculus complex (Diptera: Tephritidae). Genetics and Molecular Biology 22, 507–510. Sherron, D.A. and Rai, K.S. (1984) Genetics of speciation in the Aedes scutellaris group (Diptera: Culicidae). 3. The genetic relationship of Aedes cooki with Aedes kesseli. Journal of Medical Entomology 21, 540–547. Shoemaker, D.D., Ross, K.G. and Arnold, M.L. (1996) Genetic structure and evolution of a fire ant hybrid zone. Evolution 50, 1958–1976. Simberloff, D. and Stiling, P. (1996) Risks of species introduced for biological control. Biological Conservation 78, 185–192. Sota, T. and Kubota, K. (1998) Genital lock-and-key as a selective agent against hybridization. Evolution 52, 1507–1513.


97

Sperling, F., Byers, R. and Hickey, D. (1996) Mitochondrial DNA sequence variation among pheromotypes of the dingy cutworm, Feltia jaculifera (Gn.) (Lepidoptera: Noctuidae). Canadian Journal of Zoology 74, 2109–2117. Stouthamer, R., Jochemsen, P., Platner, G.R. and Pinto, J.D. (2000) Crossing incompatibility between Trichogramma minutum and T. platneri (Hymenoptera: Trichogrammatidae): Implications for application in biological control. Environmental Entomology 29, 832–837. Swanson, J., Lancaster, M., Anderson, J., Crandell, M., Haramis, L., Grimstad, P. and Kitron, U. (2000) Overwintering and establishment of Aedes albopictus (Diptera: Culicidae) in an urban La Crosse virus enzootic site in Illinois. Journal of Medical Entomology 37, 454–460. Taylor, D.B. (1990) Genetics of interspecific hybridization in the Triseriatus and Zoosophus groups of Aedes (Protomacleaya) (Diptera: Culicidae). Annals of the Entomological Society of America 83, 1181–1191. Taylor, D.B. and Craig, G.B. (1985) Unidirectional reproductive incompatibility between Aedes (P.) berlandi and Aedes (P.) hendersoni (Diptera: Culicidae). Annals of the Entomological Society of America 78, 769–774. Thelwell, N.J., Huisman, R.A., Harbach, R.E. and Butlin, R.K. (2000) Evidence for mitochondrial introgression between Anopheles bwambae and Anopheles gambiae. Insect Molecular Biology 9, 203–210. Thorpe, K.W. (1984) Seasonal distribution of Trichogramma (Hymenoptera, Trichogrammatidae) species associated with a Maryland soybean field. Environmental Entomology 13, 127–132. Tilmon, K.J., Wood, T.K. and Pesek, J.D. (1998) Genetic variation in performance traits and the potential for host shifts in Enchenopa treehoppers (Homoptera: Membracidae). Annals of the Entomological Society of America 91, 397–403. Umphrey, G.J. and Danzmann, R.G. (1998) Electrophoretic evidence for hybridization in the ant genus Acanthomyops (Hymenoptera: Formicidae). Biochemical Systematics and Ecology 26, 431–440. Via, S. (2001) Sympatric speciation in animals: The ugly duckling grows up. Trends in Ecology and Evolution 16, 381–390. Walker, P., Leather, S.R. and Crawley, M.J. (2002) Differential rates of invasion in three related alien oak gall wasps (Cynipidae: Hymenoptera). Diversity and Distributions 8, 335–349. Walter, H. and Lieth, H. (1967) Climate Diagrams – World Atlas. Gustav Fischer, Jena, Germany. Willett, C.S., Ford, M.J. and Harrison, R.G. (1997) Inferences about the origin of a field cricket hybrid zone from a mitochondrial DNA phylogeny. Heredity 79, 484–494. Wood, T.K. and Guttman, S.I. (1982) Ecological and behavioral basis for reproductive isolation in the sympatric Enchenopa binotata complex (Homoptera: Membracidae). Evolution 36, 233–242. Wood, T.K., Tilmon, K.J., Shantz, A.B., Harris, C.K. and Pesek, J. (1999) The role of host-plant fidelity in initiating insect race formation. Evolutionary Ecology Research 1, 317–332. Yara, K., Yano, E., Sasawaki, T. and Shiga, M. (2000) Detection of hybrids between introduced Torymus sinensis and native T. beneficus (Hymenoptera: Torymidae) in central Japan, using malic enzyme. Applied Entomology and Zoology 35, 201–206. Young, C.W., Onore, G. and Proano, K. (1999) First occurrence of Tipula (Tipula) oleracea Linnaeus (Diptera: Tipulidae) in the New World, with biological notes. Journal of the Kansas Entomological Society 72, 226–232. Yukawa, J. (1996) Identification of paedogenetic gall midge, Mycophila speyeri (Diptera: Cecidomyiidae) and possibility of accidental introduction to Japan. Japanese Journal of Applied Entomology and Zoology 40, 135–143. Zhu, Y.C., Burd, J.D., Elliott, N.C. and Greenstone, M.H. (2000) Specific ribosomal DNA marker for early polymerase chain reaction detection of Aphelinus hordei (Hymenoptera: Aphelinidae) and Aphidius colemani (Hymenoptera: Aphidiidae) from Diuraphis noxia (Homoptera: Aphididae). Annals of the Entomological Society of America 93, 486–491.

6

Assessing the Establishment Potential of Inundative Biological Control Agents

Guy Boivin,1 Ursula M. Kölliker-Ott,2 Jeffrey Bale3 and Franz Bigler2 1Horticultural

Research and Development Center, Agriculture and Agrifood Canada, 430 Boul. Gouin, St-Jean-sur-Richelieu, Québec, J3B 3E6 Canada (email: [email protected]; fax number: +1-450-346-7740); 2Agroscope FAL Reckenholz, Swiss Federal Research Station for Agroecology and Agriculture, Reckenholzstrasse 191, 8046 Zürich, Switzerland (email: [email protected]; [email protected]; fax number: +41-44-377-7201); 3School of Biosciences, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK (email: [email protected]; fax number: +44-121-414-5925)

Abstract Establishment of exotic natural enemies in the area of release is not a desirable attribute in inundative releases as it increases the risks of non-target effects on native species. To evaluate the risks of non-target effects, this chapter focuses on factors which may limit the establishment of introduced natural enemies, either for a season or permanently. From a risk assessment perspective, the risk associated with the release of a species with seasonal persistence capacity is limited in time. The establishment of natural enemies in a novel habitat depends on several factors, some abiotic and some biotic. Among the abiotic factors, climate is a major factor. Temperature and humidity, especially when soil moisture is considered in species that spend part of their development in the soil, are the components of weather that have the major impact on the survival and establishment of exotic species. Biotic factors, and especially the occurrence of alternate host/prey, also play an important role in the probability that an organism will become established. We describe in this chapter the methods that should be used to assess the probability that exotic natural enemies can become established, based on these factors. We recommend first evaluating to what extent temperature may limit establishment. Only where the risk of establishment based on thermal requirements is determined to be higher than ‘insignificant’, should the availability and suitability of host or prey for overwintering in the nontarget habitat or the impact of humidity be investigated.

Introduction In contrast to classical biological control or inoculative releases, the ability of an exotic natural enemy to establish in the area of 98

release is not a desirable attribute in inundative releases. Establishment of an introduced organism increases the risks of non-target effects on native species. Such risks include displacement of native preda-


Establishment Potential of Inundative BCA

tor or parasitoid species and a decrease in the population density of native species used as prey or host. To evaluate the risks of non-target effects, this chapter focuses on factors which may limit the establishment of introduced natural enemies. One of the risks that has to be assessed before releasing a biological control agent is the potential for the establishment of a natural enemy in areas where it is not indigenous. For example, a risk of establishment of exotic natural enemies is present if individuals escape from greenhouses in which inoculative or inundative releases are made. Mass release of predators or parasitoids in the field may also result in the establishment of these species, either for a season or permanently. Seasonal persistence is the survival and reproduction of a species throughout one season, with seasonally occurring conditions preventing further survival. Inability to overwinter due to the occurrence of low temperatures is probably the most frequent reason for failure to establish long-term populations. Permanent establishment is the survival of a species for several years. From a risk assessment perspective, the risks associated with the release of a species with seasonal persistence capacity is limited in time. If negative effects are found after the release, these effects will last for only one season. For species with the capacity to establish permanently, any damage will also be permanent. The risk factors linked to these two types of establishment should therefore be different. The establishment of natural enemies in a novel habitat depends on several factors, some abiotic and some biotic. Among the abiotic factors, climate is a major factor, and it has long been advocated that climate matching of the recipient system and the native range should help predict dispersal and potential geographic spread (Louda et al., 2003; Cock et al., Chapter 12, this volume). Temperature is the component of weather that has the major impact on the survival and establishment of exotic species. Humidity can also be a limiting factor, especially when soil moisture is considered in species that spend part of their development in the soil. These abiotic

99

factors are well characterized in most areas and the response of natural enemies to the conditions expected in the area of release can be tested. In fact, the impact of factors such as temperature and humidity should be among the first ones to be tested, especially in areas where extreme conditions are expected. The importance of abiotic factors, and mostly temperature, on the probability of establishment of a natural enemy is highlighted by the fact that most of the successes in classical biological control programmes have occurred in warm climates (DeBach, 1964). Data on temperature and humidity can be used to predict the distribution of species that have been previously introduced. However, in situations in which outdoor establishment is undesirable, such as with escapes from greenhouses, it is now apparent that climate matching between native and introduced ranges may not be a sound basis for predicting long term survival (Hart et al., 2002a,b), and more comprehensive analyses of thermal tolerance are required. Biotic factors also play an important role in the probability that an organism will become established. The occurrence of alternate host/prey, the presence of competitors or natural enemies and, finally, access to other food sources, are all factors that are important in the capacity for an organism to establish itself in an area. These factors are more difficult to assess than the abiotic factors and, for many organisms, the information available from both the area of origin and the area of introduction is often lacking or patchy. Finally, the interaction of abiotic and biotic factors will act together on both the host/prey and the natural enemy. The probability of establishment of organisms in temperate climates is affected both by mortality due to climatic extremes and by the difficulty in adjusting and synchronizing their lifecycle with that of their host, especially when these hosts enter quiescence or diapause stages during the season when extremes are reached (Bale, 1991a). This ability to synchronize with their host/prey is especially critical in specialist natural enemies because they cannot use alternate

100

G. Boivin et al.

hosts/prey to either survive until emergence or sustain populations between generations of the host/prey. The current trend toward using specialist species is sound, both from an environmental perspective and because it decreases the probability of establishment. The use of specialist species is thus generally preferable, and is particularly useful for greenhouses, where their external establishment is undesirable. We will consider in this chapter only factors that could prevent establishment of a natural enemy. Factors that affect the efficacy of the introduced organism but that will not prevent a species from establishing will not be covered. The occurrence of food sources is one such factor. The availability of an adequate food source is known to increase parasitoid efficacy but it may be impossible to demonstrate that no food source is available and that this absence will render establishment impossible. Other factors not likely to be decisive in preventing establishment include competition with other natural enemies and the presence of predators or hyperparasitoids. Therefore, we will concentrate on temperature and humidity among the abiotic factors, and on the presence of host/prey in the area of release among biotic factors. In this chapter we will briefly describe the methods that can be used to quantify these factors and ultimately determine the risk of establishment.

Factors Preventing Establishment Abiotic factors The methods that can be used to measure abiotic factors that may limit the establishment of exotic natural enemies are summarized in Table 6.1. The table includes a short description of the methods, the information gained from the experiments, and lists the equipment needed to perform the tasks. Temperature Temperature influences the probability of establishment of insect natural enemies and

the range of their distribution once established. Both low and high temperatures are to be considered, although the meaning of ‘high’ and ‘low’ will vary according to the area of origin of the organism. Temperature can affect the probability of establishment either through the thermal budget of an organism or through direct mortality caused by exposure to low or high temperature. The thermal budget refers to the accumulation of day degrees necessary to complete a generation and can be used to assess the number of generations that are theoretically possible per year. When the total number of day degrees available in an area is either below the minimum needed for a generation, or is such that at the end of summer the organism is in a stage where it cannot survive winter, establishment is unlikely. Direct mortality attributable to temperature can be due either to short exposure to lethal temperatures or to prolonged exposure to sub-optimal temperatures that become lethal over time. Most insect species have a thermal optimum at which survival and development are normal. As the temperature decreases, the insect eventually enters a sub-optimal zone, where mortality will occur after a certain time at that temperature. Below that temperature, the insect enters the temporary cold stupor zone, where vital functions such as feeding and mating are strongly reduced. Finally, it enters the chill coma, where movement becomes slower and eventually stops (Vannier, 1994). When the temperature of the insect body falls below 0°C, the haemolymph (or other tissues) eventually freezes at the supercooling point. A similar gradation can be found as the temperature increases. The insect will enter the temporary heat stupor zone, where it shows loss of coordination and short episodes of lethargy. As the temperature increases, the insect becomes motionless in the thermostupor zone (heat coma). The insect eventually dies when the temperature reaches the upper thermal death point (Vannier, 1994). As for all invertebrates, the rate of development of insects varies with temper-


ature. Several models have been used to describe the relationship between temperature and the developmental rate of insects but most have in common that above a certain temperature, the base temperature, day degrees start to accumulate. Above the base temperature, rate of development increases gradually, often with a positive slope, up to a certain temperature where the slope becomes negative. This sigmoid curve reaches a peak temperature at which the rate of development is at a maximum. Above this temperature, rate of development decreases, often quite rapidly, down to the point where no development occurs at all (Fig. 6.1). Two factors are important from the perspective of establishment. The first is the base temperature, as it is needed to calculate the accumulation of day degrees, and the second is the thermal budget, which is the number of day degrees necessary for an organism to complete a generation. Exposure to low temperature can kill insects either by freezing or by cumulative cold damage, without freezing. Two strategies have been described by which insects survive at low temperature: freeze tolerance and freeze intolerance. Freeze-tolerant species generally freeze at relatively high temperatures (above ⫺10°C) and can recover when they thaw. In these species,

101

the presence of polyols is common, and these products protect the frozen tissues from frost damage. The majority of overwintering insects are freeze-intolerant (Bale, 1991b) and are killed at the moment they freeze at the supercooling point. These species must avoid freezing, either by behavioural or physiological adaptations. Behavioural adaptations include selection of protected overwintering sites and migration away from the geographical area where temperatures lower than the supercooling point occur. Physiological adaptations involve emptying the gut to avoid the presence of ice-nucleating particles and synthesis of cryoprotectants. The use of the supercooling point to assess the cold-hardiness of a species, and therefore its probability of establishment in an area based on the lowest temperature occurring in this area, is relevant only for species where winter mortality occurs predominantly at or close to the freezing temperature of the insect. A well-known example is the autumnal moth, Epirrita autumnata (Borkhausen) (Lepidoptera: Geometridae), that occurs on mountain birch in northern Europe. The overwintering eggs of this species have a mean supercooling point of ⫺35.9°C and egg survival correlates well with the lowest tempera-

Fig. 6.1. Example of a temperature response curve (based on the equation of Brière et al., 1999).

102

G. Boivin et al.

tures during winter (Tenow and Nilssen, 1990), an indication that for this species the supercooling point is an accurate measure of cold hardiness. Classification of insects as freeze-tolerant or freeze-intolerant takes into account only freezing as a cause of death. Although this factor is relevant for species in temperate or subarctic climates that survive well at low temperatures above their supercooling point, for most species, mortality caused by low temperature occurs at temperatures much higher than the supercooling point (Bale and Walters, 2001). Exposure at temperatures above the supercooling point induces mortality, following cumulative-cold injuries, that is proportional to both the temperature and the duration of the exposure. This mortality appears to result from membrane phase transitions and protein conformational changes at low temperature (Sinclair et al., 2003). For these species, the supercooling point is an unreliable index of cold-tolerance and the impact of sub-freezing temperature must be measured. In tropical species, cumulative-cold damage may occur even at temperatures above 0°C. While the damage caused by brief periods of chill coma is readily reversible, long periods of low temperature may prove fatal (Denlinger and Lee, 1998). Some of the damage caused by low-temperature exposure can be reduced if the organism is exposed to pulses of higher temperature. These periods at higher temperature could enable insects to regenerate certain energy resources or cryoprotectants that are progressively depleted at low temperature (Denlinger and Lee, 1998). METHODS OF ASSESSING TEMPERATURE EFFECTS.

When the effect of low-temperature exposure is measured, the timing of the observation is important. Mortality can increase progressively when the individuals are returned to favourable conditions and therefore an early mortality assessment can underestimate the effect of the cold exposure (Bale, 1991a). In addition, the effect of cold exposure can be apparent only at a later stage of development, either when the

individual may die or when sub-lethal effects that reduce the fitness of the individual appear, such as reduced fecundity. Depending on the climate of the areas of origin and on the introduction of a natural enemy, different cold-related indices have been proposed to assess the probability of establishment of alien species in the UK (Bale and Walters, 2001). When these indices are used, it can be informative to test both the natural enemy to be introduced and a native related species. The native species is known to survive in the area of introduction and results may differentiate between this species and the exotic species, in which case an assessment can be made of the likelihood of establishment of the non-native species. If the climate in the area of origin differs greatly from the area of introduction, then establishment of the introduced organism in the release area may be unlikely. Information on the biology (e.g. overwintering stages, diapause characteristics), ecology (e.g. overwintering sites, migratory performance) and thermal requirement (e.g. base temperature, thermal budget) of the introduced organisms (if available) may also help to determine their potential to establish in a specific area. This information should also cover all aspects related to the capacity of the organism to adapt to a new environment, including its response to humidity. For example, the ability of an organism to enter diapause in its area of origin is likely to increase the chance of establishment in the area of introduction. LITERATURE STUDY.

DEVELOPMENTAL THRESHOLDS. The lower developmental threshold, or base temperature, is the temperature below which no development occurs. This temperature is established by obtaining the rate of development of the organism at different temperatures and then calculating the temperature at which the development rate is zero. No development will occur during periods where the maximum daily temperature is below the base temperature. The upper developmental threshold


is the temperature above which no development occurs. These thresholds can be used to calculate the number of day degrees required for a species to complete a generation. Day degrees have been used quite extensively to express insect development over the growing season. Most published estimates of day degree accumulation include the lower developmental threshold temperature (Tbase), but only a few include the optimum or higher developmental threshold temperature (Tsup). The development rate of an organism (i.e. 1/day) in relation to temperature is generally linear over the optimum temperature range but becomes curvilinear close to the Tbase and Tsup (Fig. 6.1). The use of a linear regression to estimate the Tbase (the intercept on the x-axis) may thus lead to important errors at temperatures close to the thermal extremes of the organism. Nonlinear equations are therefore preferable to express development rate as a function of temperature. These can be classified into three broad categories based on their capacity to determine Tbase and Tsup as summarized below: (i) direct estimations of Tbase and Tsup (e.g. Brière et al., 1999), (ii) indirect estimations of Tbase and Tsup (e.g. Duthie, 1997) and (iii) indirect estimations of Tbase and direct estimation of Tsup (e.g. Lactin et al., 1995). In most cases, these equations provide accurate estimates of optimum temperature when appropriate data of insect development as a function of temperature are available. The thermal budget is the number of day degrees required by a species to complete a generation. The base temperature (developmental threshold) must be established before this index can be calculated. When the annual accumulation of day degrees in an area is below that required to complete a generation, a natural enemy will not be able to establish. In addition, if the day degree accumulation permits the development of partial generations, the impact could be either to reduce the size of the population or even to pre-

THERMAL BUDGET.

103

vent its establishment. When it is of interest to consider partial development (i.e. number of day degrees necessary to complete diapause or specific stages of the life cycle), a similar approach is used and the accumulation of day degrees recorded each day until the desired phase of development is completed. POINT. The supercooling point (SCP) is the temperature at which an individual freezes. For freeze-intolerant species, death occurs at or above this temperature. Although is it recognized that death can occur at temperatures much above the supercooling point, this temperature is still relevant, especially for species originating from cold areas. Also, the supercooling point indicates the temperature above which the incidence of prefreeze mortality can be investigated. The supercooling point is determined by detecting the small increase of temperature resulting from the release of latent heat when body water freezes. The organism is cooled at a constant rate (typically 1°C/min) while its temperature is continuously recorded. Microthermocouples, of type T (copper-constantan) or K (chromel– alumel), are normally used and temperature recording done on either a paper chart or a data logger (Hance and Boivin, 1993). Preparation of the organism is also critical as age, feeding status, surface particles or water film can modify the freezing temperature through the presence of ice nucleators or ice crystals. In addition, although a cooling rate of 1°C/min is usually used, this rate is much higher than occurs in natural situations (Sinclair, 2001). Variation in cooling rate has little effect on the SCP, but may modify the ability of insects to survive the freezing event. Cooling at a constant rate can be achieved by apparatus using a watercooled Peltier effect module linked to an electronic control unit (Bale et al., 1984; Panneton et al., 1995). Such systems control the temperature within ± 0.2°C. It is also possible to achieve an approximately linear decrease of temperature by placing the insect in an insulated container within SUPERCOOLING

104

G. Boivin et al.

a large freezer at ⫺30°C (Hance and Boivin, 1993). However, the decrease in temperature tends to become curvilinear as the temperature within the insulated container approaches the temperature of the freezer. Care must be taken to test the developmental stage of the organism that will normally overwinter. Organisms that overwinter in diapause must also be in diapause when tested, as diapause often changes the supercooling point or, more importantly, the cold tolerance. LETHAL TEMPERATURE 50 (LTEMP50). The LTemp50 is the temperature at which 50% of a population dies. However, in some species, the determination of this temperature depends on the duration of exposure. As exposure time decreases, the LTemp50 may be closer to the supercooling point. Since the purpose of estimating the LTemp50 is to identify the temperature at which 50% of a population are killed, organisms are usually cooled at 1°C/min and exposed to a series of decreasing subzero minimum temperatures for 1 min with mortality assessed 24 and 48 h after exposure. The organism should be placed in a closed container, to avoid damage through desiccation and to buffer the decrease and increase in temperature that will occur at both the beginning and the end of the experiment. The range of temperatures to be tested should be based on the low temperatures experienced by the species in its natural habitat, but should be above the measured supercooling point. Ideally the organism should be in a state compatible to that which overwinters, i.e. for many species low-temperature exposure occurs when the insect is in forms of dormancy, such as diapause or quiescence. The data are obtained as percentage mortality at decreasing temperatures and are then analysed by probit or logit to derive an estimate of the LTemp50. It is important to recognize that this index assesses mortality but does not take into account any sublethal effects such as reduction of longevity, fecundity or modification of behaviour that may occur in surviving insects.

LETHAL TIME 50 (LT50). This index is based on the duration of exposure at a given temperature sufficient to cause 50% mortality in a population. In a sense it is the reverse of LTemp50 but with constant exposure temperatures. Choosing the temperatures to be tested may, however, prove difficult. These temperatures should be chosen so as to be similar to the low temperatures likely to be experienced in the area where the natural enemy will be released. Several containers containing individuals to be tested are placed at the selected temperatures and, at intervals, replicate samples are removed, placed at a standard temperature (i.e. 20°C or 25°C) and survival of the individuals is assessed. Survival should be assessed 24 h and 48 h after removal from the low-temperature environment, and the data analysed by probit or logit. Although low temperature can affect longer-term survival or reproduction (Bale, 1991a), experiments that can detect such effects are long and costly and unlikely to be performed on a routine basis.

If local regulations permit, outdoor cages should be used to assess winter survival. Survival of a species under semi-natural conditions in an outdoor cage is as close an approximation of natural conditions as can be obtained. Multiple cages should be used as independent replicates. However, it should be noted that conditions within the cages may differ from the habitat where the release will be made, and that atypical winters may over- or underestimate mortality. Temperature recording inside the outdoor cages may help to explain unexpected results. The results from outdoor cage tests can be used to verify the findings obtained in the laboratory tests and the predictions on winter survival and hence the likelihood of permanent establishment. As an example, outdoor cage studies were performed with Trichogramma brassicae Bezd. to assess overwintering in six different host eggs under natural conditions in Switzerland (Babendreier et al., 2003). A summary of the methods discussed above is tabulated in Table 6.1.

OUTDOOR CAGE TESTS.


105

Table 6.1. Summary of methods that can be used to assess the likelihood of establishment of an introduced natural enemy based on abiotic factors. Assessment

Description of method

Temperature Literature study

Climate matching between the area of origin and introduction; ecological information on the organism Developmental The organisms are reared threshold at different temperatures Thermal budget The organisms are reared at different temperatures Supercooling point The organism is cooled at constant rate (1°C/min) while its temperature is recorded

Lethal temperature Mortality at 24 or 48 h at a specific temperature

Lethal time

Field cage tests Humidity

Equipment needed

Information gained

No special equipment needed

If climates differ then establishment is unlikely; thermal requirements of the organism (if available)

Climatic chamber with controlled temperature Climatic chamber with controlled temperature Microthermocouples, temperature recording either on a paper chart or a data logger, watercooled Peltier effect module controlled by an electronic control unit or insulated container within a large freezer at ⫺30°C Climatic chamber with controlled temperature

Base temperature

Mortality after certain lapse of time at a specific temperature Mortality

Climatic chamber with controlled temperature Field or outdoor cages

Assess survival, fecundity etc. at different temperatures and humidities

Climatic chamber with controlled temperature, desiccators with salt solutions

Humidity While temperature is usually the predominant abiotic factor preventing establishment, humidity may influence long-term survival of introduced species if the area of origin and introduction differ in humidity conditions. Evaluating performance and survival at low and high humidities may help to predict limitations to the establishment of introduced natural enemies. While temperature directly influences development and survival, humidity effects may be less pronounced since organisms have the ability to regulate their body water content to some extent. Hadley (1994) describes the mechanisms used by

Day degree requirement to complete one generation Cold tolerance, acclimation ability

Cold tolerance as a function of exposure temperature; acclimation ability Cold tolerance as a function of exposure time; acclimation ability Ability to overwinter under near field conditions Tolerance to different temperature/humidity combinations

terrestrial arthropods to maintain water balance. Water is gained by drinking, eating, metabolizing food items and absorbing vapour from the atmosphere. For the majority of species, water in the diet is sufficient to balance losses. The processes by which water is lost include cuticular and respiratory transpiration, passive diffusion from oral and anal openings and water loss associated with excretion. Cuticular transpiration constitutes the major avenue of water loss despite the presence of a highly waterproofed integument in most species. In terrestrial arthropods, the epicuticle provides the principal barrier to water loss. Quantitative differences in cuticular lipids

106

G. Boivin et al.

may contribute to the lower water loss rates and hence increased desiccation resistance of some species (Hadley, 1994). For most species, absolute water loss increases at lower humidities as a result of the lower saturation of the surrounding air. In contrast, the calculated permeability of the cuticle (corrected for saturation deficit) often increases as humidities rise. This may facilitate water loss and thus prevent the arthropod from becoming overhydrated (Hadley, 1994). If the humidity conditions become unfavourable for species survival, individuals of the free-living stages may be able to reduce or avoid dehydration stress by clustering (Yoder and Barcelona, 1995; Yoder and Smith, 1997), by moving into microhabitats more suitable for survival, or by using avoidance behaviours such as burrowing and nocturnal activity. The desiccating conditions present on the surface can largely be avoided by moving a few centimetres into the soil, or by restricting the surface activity to night-time hours when humidities are higher (Hadley, 1994). Humidity can also be higher in aboveground microshelters, e.g. condensation on the undersides of rocks may provide water. Soil moisture may influence survival of natural enemies with life stages living in the soil. The diameter of pores between soil particles decreases with increasing depth. A portion of the pore system is often filled with water that accumulates from rainfall or rises as capillary groundwater. The remainder of the pore system contains air that is saturated with water vapour (Eisenbeis and Wichard, 1987). While mobile life stages may undertake horizontal and vertical migrations to retain access to moisture during hot and dry periods, immobile life stages (i.e. eggs, pupae) may be affected by changes of humidity within soil pores. A range of relative humidities can be generated using saturated salt solutions (Winston and Bates, 1960) or glycerol–water mixtures (Johnson, 1940). Calcium sulphate (Drierite) provides 0% relative humidity. The relative

METHODS OF ASSESSING HUMIDITY EFFECTS.

humidities, as well as the temperatures chosen for the experiment, should reflect the conditions in the non-target habitat when humidity conditions become extreme. Desiccation tolerance is influenced by a variety of factors including age, sex and life history stage (Eckstrand and Richardson, 1980; Lamb, 1984; Hadley, 1994). The free-living stages may be more susceptible to humidity extremes than stages protected within hosts. These facts should be taken into account when selecting individuals for the test. Experiments assessing humidity effects on introduced species should include a taxonomically related resident species as a control (as was done for temperature in Bale and Walters (2001)) and for comparison of results. If, for example, the local species dies at the humidity extremes naturally occurring in the non-target area, then protected microhabitats with more favourable environmental conditions may be available for survival during unfavourable environmental conditions. In order to assess if local humidity conditions can cause lethal or sub-lethal effects on the exotic natural enemy, the introduced species may be reared at or exposed to different temperatures and humidities in small chambers (incubators). For economic reasons, not all temperature/humidity combinations that occur in the area of release can be tested in practice, hence we propose that the most current extremes of temperature and humidity should be identified (from climate records) and the relevant stages tested under these conditions. The parameters used to assess humidity effects may include egg hatch rate, development time, pre-imaginal survival, fecundity or longevity.

Biotic factors Host/prey Among other factors, establishment of an exotic natural enemy depends on the availability and suitability of hosts or prey and their spatial and temporal synchronization


with the introduced organism. First, a list of potential hosts or prey that are taxonomically related and occur in habitats similar to that in which the new agents would be released has to be established. Then, it has to be determined which of these hosts or prey match in space and time with the introduced organism, e.g. do they occur at the same altitude and time as the natural enemy (for more details see Kuhlmann et al., Chapter 2, this volume). Next, the acceptance and suitability of the remaining species can be assessed in host specificity tests (for more details see van Lenteren et al., Chapter 3, this volume). Permanent establishment is only possible if overwintering hosts are available. The indigenous alternate hosts or prey may not be adequate for overwintering of the natural enemy, either because of their intrinsic capacity to survive harsh conditions at the stage during which they are attacked, because of a lack of synchronization with the exotic natural enemy, or because its type of dormancy is unsuitable for the parasitoid or predator. When the host is not at a suitable stage for winter survival at the time the natural enemy is preparing for winter, no permanent establishment will occur. This is the case for several Trichogramma species that need diapausing eggs of their host to survive winter (Boivin, 1994). METHODS OF ASSESSING HOST/PREY EFFECTS.

The methods for obtaining a list of potential hosts or prey and for testing their suitability for survival and reproduction of the natural enemy are described by Kuhlmann et al. (Chapter 2, this volume) and by van Lenteren et al. (Chapter 3, this volume). After a list of potential hosts or prey has been established, the availability and suitability for overwintering of the introduced organism and the abundance and ecological significance of these species can be determined. The methods described in the temperature and humidity sections should be used to assess survival of the exotic natural enemy in the overwintering stage of the potential hosts or prey. Care must be taken to use the appropriate developmental stage

107

of both the host and the natural enemy and to make sure that they are in dormancy if they overwinter in this condition.

Case Studies Using Temperature to Assess the Establishment Potential of Non-native Biological Control Agents in the UK The protocols for the practical assessment of establishment potential of non-native invertebrate biological control agents can be developed from a theoretical analysis of the requirements of such species when introduced into a new environment. In simple terms, if a non-native species is introduced into a greenhouse ecosystem, in a region with a winter season, for outdoor establishment to occur, any escaping individuals will require a combination of (i) a thermal budget above the developmental threshold sufficient to complete at least one generation per year, (ii) one or more life stages able to survive at low temperature, (iii) the ability to enter quiescent or diapause states, and (iv) sources of host or prey. In the context of this chapter, the interrelationships between temperature and development and winter survival have recently been investigated in a number of insect and mite greenhouse biological control agents introduced into the UK over the past 15 years. The work was conducted to seek ecophysiological explanations for the ‘unexpected’ establishment of some introductions and, in turn, to develop experimental approaches that could be used to assess the establishment potential of candidate species under current or future consideration for import and release. The predatory mite Neoseiulus (Amblyseius) californicus (McGregor) (Acari: Phytoseiidae) was first released in UK greenhouses in 1991 and within ten years was reported to have established wild populations in areas close to release sites. The predatory mirid Macrolophus caliginosus Wagner (Heteroptera: Miridae) was released in 1995 and has been observed outside of greenhouses at different times of the year, though establishment

108

G. Boivin et al.

has not yet been confirmed. Two other species, the parasitoid Eretmocerus eremicus Rose and Zolnerowich (Hymenoptera: Aphelinidae) and the predatory ladybird Delphastus catalinae (Horn) (Coleoptera: Coccinellidae), are both licensed for release in the UK, with no reports of winter survival or establishment. Currently, another predatory mite, Typhlodromips montdorensis (Schicha) (Acari: Phytoseiidae), is being considered as a candidate species for release. A range of experimental procedures have been applied to these species to determine their developmental threshold temperature, thermal budget (day degree) requirement per generation, potential annual voltinism, cold tolerance (freezing temperature, lethal temperatures and times) and acclimation ability, and response to diapause-inducing cues and winter field survival. Using M. caliginosus as a case study to exemplify this approach (Hart et al., 2002a), the threshold temperature for development from egg to adult was estimated to be 8.4°C (simple linear regression) and 7.7°C (weighted linear regression) with thermal budget requirements of 472 and 495 day degrees, respectively, above the threshold. Analysis of developmental data for a range of species suggests that a line derived from simple linear regression does not fit closely with data points at the lowest experimental temperature, sometimes resulting in an inaccurate estimate of the threshold temperature; in most cases, this problem can be overcome by the application of weighted linear regression. For some species it may be valuable to examine differences in threshold temperatures between different life cycle stages to identify possible ‘rate-limiting’ stages. For example, the threshold temperatures for the egg and nymphs of M. caliginosus calculated by weighted linear regression are 8.7° and 7.2°C, respectively. Estimates of the developmental temperature and thermal budget can then be related to climate records for any intended release site to determine the likely annual voltinism. For instance, in the Midlands area of the UK, the annual number of day

degrees above the developmental threshold of 7.7°C varied from 1059 to 1347 (mean 1253) over the ten-year period from 1991 to 2000, indicating that in all but one year, M. caliginosus would have been able to complete two (but never three) generations. By inspection of the monthly totals of available day degrees it is possible to determine whether development is restricted to the summer months, or can proceed through winter. Also, the development data for a particular species or strain can be related to climate records for any release site, in different countries or regions of the world. Cold-tolerance assessments were made on two age groups: first/second instar nymphs and fifth instar/adults. The likelihood of winter survival is increased if individuals escaping from greenhouses are able to acclimate at lower temperatures. In most insects capable of an acclimation response, significant changes in one or more indices of cold tolerance are usually detectable after seven to ten days at 5–10°C. Acclimation regimes are therefore intended to detect the ability to acclimate rather than to produce ‘fully acclimated winterhardy’ populations. The mean supercooling points of the two tested age groups of M. caliginosus with and without acclimation at 10°C for seven days in a 12:12 LD cycle varied from ⫺19.0 ± 0.6° to ⫺20.3 ± 0.3°C, with no significant difference between the groups. Supercooling points of many insects lie in the range of ⫺15° to ⫺25°C, but low freezing temperatures are not, in isolation, a reliable indicator of cold tolerance. Lethal temperatures (LTemp) of M. caliginosus were calculated by cooling replicate samples of the four treatment groups at 0.5°C/min to temperatures between ⫺5° and ⫺19°C (mean supercooling point of the ‘least’ cold-hardy age group), with exposure of 1 min at the minimum temperature. Probit analysis of the data provides estimates of the temperatures required to kill given proportions of each treatment group, typically, 10, 50 and 90%. The lethal temperatures of the two age groups were similar, with no acclimation response and LTemp50 values consistently around ⫺15°C,


indicating some ‘pre-freeze’ mortality. Estimates of supercooling points and lethal temperatures often show values that are lower than the minimum temperatures likely to be experienced in the region or country of release; for instance, temperatures of ⫺15°C or lower rarely occur in the UK. However, the supercooling point and LTemp50 are both indices that are measured after very brief exposures. It is likely, therefore, that estimates of the duration of survival at less severe temperatures will provide a more informative guide to survival under field conditions. When replicate samples of the different M. caliginosus age and acclimation groups were exposed at ⫺5°, 0° and 5°C for increasing periods of time, the lethal time (LT) values increased at the higher temperatures, with LT50 values at 5°C of around 20–30 days for the different groups. These extended exposure experiments are more realistic in terms of natural conditions, but other factors, such as starvation, may affect the observed survival. During intermittent periods of higher winter temperatures, natural enemy species may search for hosts and prey and extend their survival. Laboratory experiments should therefore include treatments that provide access to prey. When the LT experiments at 5°C were repeated with greenhouse whitefly, Trialeurodes vaporariorum (Westwood) (Homoptera: Aleyrodidae), added as prey, 50% survival time of ‘fed adults’ increased from 30 to 50 days, with 10% still alive after 75–80 days. Whilst these ‘time’ experiments approximate more to field situations, insects and mites will usually be subject to fluctuating temperatures. There are now many examples where duration of survival is increased when insects, kept at constant low and stressful temperatures (often in chill coma), are periodically transferred to higher ‘recovery’ temperatures, and thus able to move and feed. For these reasons it is possible that laboratory exposures, even at 5°C, will underestimate the field survival of insects and mites originating from Mediterranean or tropical climates, especially during mild winters. Assessment of survival in the field is there-

109

fore an essential component in the risk assessment of establishment potential. Biological control agents may escape from greenhouse environments at any time of the year, and therefore encounter conditions that may be temporarily favourable or more or less immediately lethal. More specifically, organisms that escape at the end of the summer will have to survive in the field for six months or longer before favourable conditions return, whereas, for those escaping in mid-winter, the cold and starvation stress will be less prolonged. Whilst this difference in the time of escape is unlikely to affect permanent establishment, it may allow a species to persist in the field until the next winter. When nymphs and adults of M. caliginosus (with no whitefly prey) were placed in the field in November and January (to represent early- and mid-winter escapes from greenhouses), there was a progressive decline in survival, with 100% mortality after 40 and 60 days for nymphs and adults, respectively, with the microhabitat temperature rarely falling below 0°C. A similar pattern was observed in the following winter, with maximum nymph and adult survival times of 40–60 days with the temperature falling to ⫺5°C on some occasions. However, in the same winter, when M. caliginosus were provided with whitefly prey, adult survival increased to 75 days, and more importantly, some nymphs developed in the field and were still alive after 200 days, the duration of a full temperate winter (Hart et al., 2002a). We conclude from these case studies that there are other considerations to take into account in the planning and interpretation of field experiments. First, if a species is able to enter a diapause state, which is usually associated with increased cold tolerance and the ability to withstand starvation, winter survival and long-term establishment is more likely to occur. The diapause trait in some source populations of N. californicus is a major contributing factor to its establishment in the UK. There is, though, a second important factor, also exemplified by N. californicus. Some insect and mite biological control agents have

110

G. Boivin et al.

rapid generation times, such that, even at lower temperatures, the individual that starts the winter will never survive until the end of winter. However, if escaping individuals can reproduce in the field, their progeny may be able to sustain the population until the following spring which, in turn, obviates the need to enter a diapause state (Hart et al., 2002b). The application of the experimental protocol described for M. caliginosus to other species provides an opportunity to investigate the combined datasets to identify laboratory indices that are reliable predictors of field survival in winter. This has been done for M. caliginosus (Hart et al., 2002a), N. californicus (Hart et al., 2002b), E. eremicus (Tullett et al., 2004), T. montdorensis (Hatherly et al., 2004) and D. catalinae. For these species, a strong correlation has been found between the LT50 at 5°C in the laboratory and the duration of winter field survival (Fig. 6.2). On the basis of this relationship, it appears that a reliable prediction of winter field survival can be

obtained from a relatively rapid laboratory assay. Clearly, this approach is likely to be attractive to biological control companies in the production of the environmental risk assessment dossier that accompanies a licence application, as it focuses limited research and development budgets on a critical range of experiments. In this respect, whilst the ‘LT50 prediction’ provides an accurate ‘retrospective’ ecophysiological explanation for the establishment success and failure of a range of species released in the UK over the past 15 years, there also some caveats to be considered at this time. First, whilst the species so far investigated are drawn from different taxonomic groups and from different trophic guilds (predators and parasitoids), it is likely that there will be some exceptions that will not conform with the emerging laboratory–field relationship described in Fig. 6.2. It is probably too early for regulatory authorities and biological control companies to rely exclusively on the laboratory LT50 to predict establishment potential; it

Fig. 6.2. Relationship between LT50 in the laboratory and field survival of the same life cycle stages of five non-native biological control agents.


would be valuable to apply the full range of approaches described in the risk assessment protocol to further species in order to gain confidence in the predictive power of the laboratory experiments. Secondly, the LT50 winter field survival relationship is not intended to predict with ‘precise’ accuracy the maximum survival times of escaped populations of non-native biological control agents. Rather, the system should be viewed as a mechanism by which to categorize candidate biological control agents into different ‘risk’ groups. Thus with reference to Fig. 6.2, E. eremicus, D. catalinae and T. montdorensis comprise a ‘low or no risk’ group, where all field populations die out in winter after approximately one month. Macrolophus caliginosus is in a ‘marginal risk’ group, where extended survival in winter could be expected, but long-term and widespread establishment may not occur. The ability to move flexibly in winter between the greenhouse and outdoor locations (as is believed to be the case with M. caliginosus) would increase the occurrence of such species outdoors. Finally, N. californicus falls into a ‘high risk’ group, where establishment is likely, attributable to both the cold-hardy diapause strains and non-diapause populations that are sufficiently cold hardy to develop and reproduce in winter, at least in a temperate climate. Of course, as emphasized at the outset, if the species is sufficiently cold hardy to survive through winter, establishment will then depend on access to host or prey. In a wider perspective, knowledge that establishment is likely to occur is in itself not a reason to prohibit the import and release of a non-native species. It is the acquisition of that knowledge that allows a rational and informed decision to be made after an appropriate evaluation of the risks and benefits.

Conclusions and Recommendations When assessing the establishment potential of an exotic natural enemy, we recommend first evaluating to what extent temperature may limit establishment. Only where the

111

risk of establishment based on thermal requirements is determined to be higher than ‘insignificant’ (for definition of ‘insignificant’ see below), should the availability and suitability of host or prey for overwintering in the non-target habitat or the impact of humidity be investigated. We suggest that experiments should start with temperature as a limiting factor for the following reasons: (i) temperature is the most likely abiotic factor to limit establishment; (ii) temperature data for the release area are often available from meteorological offices and their acquisition requires minimal effort; (iii) testing for the temperature requirements and limitations is simpler than testing for host or prey specificity; (iv) humidity alone is seldom a limiting factor; most often it acts together with temperature; (v) other biotic factors such as food sources (pollen, nectar, etc.), or competition with other natural enemies, have rarely been reported to be the sole factors preventing establishment; and (vi) a combination of thermal budget, lethal temperature (LT50) and outdoor cage experiments may provide reliable data with adequate effort for predicting establishment. However, outdoor cage tests with exotic natural enemies may require ‘contained release’ licences from the national regulatory authority, which may not be granted. As Hart et al. (2002a,b) and Tullett et al. (2004) have shown, LT50 by itself may be a reliable predictor of field survival (Fig. 6.2). However, more tests are needed to confirm the power of prediction using LT50 in isolation from other experiments. Based on the comparison of the thermal requirements and tolerances of the exotic natural enemy to the temperature in the area of introduction, the likelihood and magnitude for establishment in non-target habitats can be categorized. The likelihood of establishment can be classified as ‘very unlikely’, ‘unlikely’, ‘possible’, ‘likely’ or ‘very likely’, and the magnitude as ‘minimal’, ‘minor’, ‘moderate’, ‘major’ or ‘massive’ (for description of risk classes see van Lenteren et al. (2003), van Lenteren and Loomans (Chapter 15, this volume, Tables 15.1, 15.2 and 15.3)).

112

G. Boivin et al.

Based on the temperature requirements of a species, the likelihood and magnitude of establishment can be assessed qualitatively and combined in a risk matrix, resulting in risk levels of ‘insignificant’, ‘low’, ‘medium’ and ‘high’ (van Lenteren and Loomans, Chapter 15, this volume, Table 15.1). The matrix can be used as a tool by the risk assessment authorities to conclude whether and to what extent temperature conditions are

limiting establishment. For organisms with seasonal persistence, the probability for permanent establishment is categorized as ‘very unlikely’ and the magnitude as ‘minimal’, therefore the risk is ‘insignificant’. If the risk of establishment is categorized as ‘low’, ‘medium’ or ‘high’, evaluation of other factors limiting establishment, such as availability, acceptance and suitability of overwintering hosts, will be needed.

References Babendreier, D., Kuske, S. and Bigler, F. (2003) Overwintering of the egg parasitioid Trichogramma brassica in Northern Switzerland. BioControl 48, 261–273. Bale, J.S. (1991a) Implications of cold hardiness for pest management. In: Lee, R.E. and Denlinger, D.L. (eds) Insects at Low Temperature. Chapman and Hall, New York, pp. 461–498. Bale, J.S. (1991b) Insects at low temperature: a predictable relationship? Functional Ecology 5, 291–298. Bale, J.S. and Walters, K.F.A. (2001) Overwintering biology as a guide to the establishment potential of non-native arthropods in the UK. In: Atkinson, D. and Thorndyke, M. (eds) Environment and Animal Development. Genes Life Histories and Plasticity. Bios, Oxford, UK, pp. 343–354. Bale, J.S., O’Doherty, R., Atkinson, H.J. and Stevenson, R. (1984) An automatic thermoelectric cooling method and computer-based recording system for supercooling point studies on small invertebrates. Cryobiology 21, 340–347. Boivin, G. (1994) Overwintering strategies of egg parasitoids. In: Wajnberg, E. and Hassan, S.A. (eds) Biological Control with Egg Parasitoids. CABI Publishing, Wallingford, UK, pp. 219–244. Brière, J.F., Pracros, P., Le Roux, A.Y. and Pierre, J.S. (1999) A novel rate model of temperaturedependent development for arthropods. Environmental Entomology 28, 22–29. DeBach, P. (1964) Biological Control of Insect Pests and Weeds. Chapman and Hall, London, UK. Denlinger, D.L. and Lee, R.E. (1998) Physiology of cold sensitivity. In: Hallman, G.J. and Denlinger, D.L. (eds) Temperature Sensitivity in Insects and Application in Integrated Pest Management. Westview Press, Boulder Colorado, pp. 55–96. Duthie, J.A. (1997) Models of the response of foliar parasites to the combined effects of temperature and duration of wetness. Phytopathology 87, 1088–1095. Eckstrand, I.A. and Richardson, R.H. (1980) Comparison of some water balance characteristics in several Drosophila species which differ in habitat. Environmental Entomology 9, 716–720. Eisenbeis, G. and Wichard, W. (1987) Atlas on the Biology of Soil Arthropods. Springer Verlag, Berlin, Germany. Hadley, N.F. (1994) Water Relations of Terrestrial Arthropods. Academic Press, New York. Hance, T. and Boivin, G. (1993) Effect of parasitism by Anaphes sp. (Hymenoptera: Mymaridae) on the cold hardiness of Listronotus oregonensis (Coleoptera: Curculionidae) eggs. Canadian Journal of Zoology 71, 759–764. Hart, A.J., Bale, J.S., Tullett, A.G., Worland, M.R. and Walters, K.F.A. (2002a) Effects of temperature on the establishment potential of the predatory mite Amblyseius californicus McGregor (Acari: Phytoseiidae) in the UK. Journal of Insect Physiology 48, 593–599. Hart, A.J., Tullett, A.G., Bale, J.S. and Walters, K.F.A. (2002b) Effects of temperature on the establishment potential in the UK of the non-native glasshouse biocontrol agent Macrolophus caliginosus. Physiological Entomology 27, 112–123. Hatherly, I.S., Bale, J.S., Walters, K.F.A. and Worland, M.R. (2004) Thermal biology of Typhlodromips montdorensis: implications for its introduction as a glasshouse biological control agent in the UK. Entomologia Experimentalis et Applicata 111, 97–109. Johnson, C.G. (1940) The maintenance of high atmospheric humidities for entomological work with glycerol-water mixtures. Annals of Applied Biology 27, 295–299.


113

Lactin, D.J., Holliday, N.J., Johnson, D.L. and Craigen, R. (1995) Improved rate model of temperaturedependent development by arthropods. Environmental Entomology 24, 68–75. Lamb, M.J. (1984) Age related changes in the rate of water loss and survival time in dry air of active Drosophila melanogaster. Journal of Insect Physiology 30, 967–973. Louda, S.M., Pemberton, R.W., Johnson, M.T. and Follett, P.A. (2003) Nontarget effects – the Achilles’ heel of biological control? Retrospective analyses to reduce risk associated with biocontrol introductions. Annual Review of Entomology 48, 365–396. Panneton, B., St-Laurent, G. and Boivin, G. (1995) Un générateur de fonction de température pour l’étude de la résistance au froid des insectes. Canadian Agricultural Engineering 37, 287–293. Sinclair, B.L. (2001) Field ecology of freeze tolerance: interannual variation in cooling rates, freeze–thaw and thermal stress in the microhabitat of the alpine cockroach Celatoblatta quinquemaculata. Oikos 93, 286–293. Sinclair, B.J., Vernon, P., Klok, C.J. and Chown, S.L. (2003) Insects at low temperature: an ecological perspective. Trends in Ecology and Evolution 18, 257–262. Tenow, O. and Nilssen, A. (1990) Egg cold hardiness and topoclimatic limitations to outbreaks of Epirrita autumnata in northern Fennoscandia. Journal of Applied Ecology 27, 723–734. Tullett, A.G., Hart, A.J., Worland, M.R. and Bale, J.S. (2004) Assessing the effects of low temperature on the establishment potential in Britain of the non-native biological control agent Eretmocerus eremicus. Physiological Entomology 29, 1–9. van Lenteren, J.C., Babendreier, D., Bigler, F., Burgio, G., Hokkanen, H.M.T., Kuske, S., Loomans, A.J.M., Menzler-Hokkanen, I., van Rijn, P.J.C., Thomas, M.B., Tommasini, M.G. and Zeng, Q.-Q. (2003) Environmental risk assessment of exotic natural enemies used in inundative biological control. BioControl 48, 3–38. Vannier, G. (1994) The thermobiological limits of some freezing intolerant insects: the supercooling and thermostupor points. Acta Oecologica 15, 31–42. Winston, P.W. and Bates, D.H. (1960) Saturated solutions for the control of humidity in biological research. Ecology 41, 232–237. Yoder, J.A. and Barcelona, J.C. (1995) Food and water-resources used by the Madagascan hissingcockroach mite, Gromphadorholaelaps schaeferi. Experimental and Applied Acarology 19, 259–273. Yoder, J.A. and Smith, B.E. (1997) Enhanced water conservation in clusters of convergent lady beetles, Hippodamia convergens. Entomologia Experimentalis et Applicata 85, 87–89.

7 Methods for Monitoring the Dispersal of Natural Enemies from Point Source Releases Associated with Augmentative Biological Control Nick J. Mills,1 Dirk Babendreier 2 and Antoon J.M. Loomans 3 1Environmental

Science, Policy and Management, 127 Mulford Hall, University of California, Berkeley, CA 94720-3114, USA (email: [email protected]; fax number: +1-510-643-5438); 2Agroscope FAL Reckenholz, Swiss Federal Research Station for Agroecology and Agriculture, Reckenholzstr. 191, 8046 Zürich, Switzerland (email: [email protected]; fax number: +41-44-377-7201); 3Plant Protection Service, Section Entomology, PO Box 9102, 6700 HC Wageningen, The Netherlands (email: [email protected]; fax number: +31-317-421701)

Abstract Mark–release–recapture (MRR) experiments are considered the best approach to use in monitoring the dispersal of natural enemies from the target environment, in an assessment of the risk of non-target impacts from augmentative releases. Starting from some general considerations of the difficulties of using MRR, we specifically address marking techniques, the design of recapture grids and the limitations imposed by different sampling strategies for the recapture of the natural enemies released. Subsequently, we describe both an exponential and a diffusion model for dispersal that can be used to analyse the time-integrated density–distance data generated from MRR experiments, pointing out the need to examine and correct the data for directionality, if possible, or to use a diffusion model with displacement when correction is not possible. The application of the exponential and diffusion models of dispersal to the estimation of dispersal distance and density, the two most important metrics to consider in a risk assessment of non-target impacts of augmented natural enemies, is also discussed. Finally, we present a case study of an inundative release of Trichogramma brassicae in a meadow in Switzerland to illustrate how the data from an MRR experiment can be fitted to a dispersal model to estimate dispersal distance and the density of dispersing individuals at different distances from the release point.

114


Methods for Monitoring the Dispersal of Natural Enemies

Introduction The risk of non-target impacts has become of increasing concern in the biological control of arthropod pests (Simberloff and Stiling, 1996; Follett and Duan, 2000; Louda et al., 2003). The focus of attention has been on understanding the host range of imported natural enemies used for the control of invasive pests, the potential for evolutionary host range expansion of imported natural enemy populations and the consequences of natural enemy impacts on non-target species at the population level (Hoddle, 2004). Although the risk of non-target impacts from imported natural enemies poses the greatest concern to ecologists and environmentalists, inundative biological control agents, or those natural enemies that are mass reared to locally inundate managed ecosystems for the control of arthropod pests (Daane et al., 2002), also have the potential to cause non-target impacts in the surrounding landscape (Lynch et al., 2001). With this in mind, an initial step in the evaluation of the environmental risks of augmentative biological control has been made by van Lenteren et al. (2003), who recently developed a risk index for the commercially available inundative control agents used in greenhouse or open-field crops in Europe. Augmentative biological control includes both the inoculative release of smaller numbers of natural enemies for season-long control of arthropod pests and the inundative release of very large numbers of mass-produced natural enemies for the rapid, but only temporary, suppression of arthropod pests (Daane et al., 2002). The natural enemies used in augmentative biological control may either be indigenous or exotic, and while all groups of arthropod natural enemies have been considered, we will focus here on invertebrate natural enemies in accordance with the OECD (2004) Guidance for Regulation of Invertebrates as Biological Control Agents (IBCAs). Insect parasitoids, insect and mite predators and

115

entomopathogenic nematodes are mass produced by an increasing number of commercial suppliers and government research institutes worldwide, and are widely available for augmentative release for the suppression of a variety of arthropod pests of managed crops and livestock production (van Lenteren, 2003). Using the approach of Hickson et al. (2000), van Lenteren et al. (2003) proposed that non-target impacts of augmentative biological control agents be assessed from the likelihood (probability) and the magnitude (consequences) of adverse effects based on the following five risks relating to the ecology of the natural enemy: (i) establishment in the target region if the natural enemy is exotic, (ii) dispersal from the target environment, (iii) host range, (iv) direct effects on non-target organisms and (v) indirect effects on other organisms in the target environment. In this chapter we will focus exclusively on dispersal from the target environment, and discuss methods used to quantify and analyse the dispersal of natural enemies from a central release point, taking Trichogramma brassicae Bezdenko as a case study.

Potential for Adverse Impacts of Natural Enemies on Non-targets from Dispersal Dispersal is the exploratory, undirected movement of individuals away from the habitat of origin (den Boer, 1990). In the context of environmental impacts of augmentative biological control, this represents the undirected movement of natural enemies away from the release site and into the surrounding landscape. In most cases, the dispersal of natural enemies will be by flight in the adult stage, as the movement of juvenile stages is restricted to a very local scale. There are two interesting exceptions, however, one in which dispersal of juvenile entomopathogenic nematodes occurs through flight of the adult

116

N.J. Mills et al.

host (Lacey et al., 1995), and a second in which dispersal of adult female parasitoids (T. brassicae) is facilitated through a phoretic association with adult female butterflies (Fatouros et al., 2005). In considering the risk of non-target impacts from the dispersal of natural enemies released for augmentative biological control, two factors of potential concern are the likely distance of dispersal and the density of natural enemies at given distances from the release point. Although dispersal distance is potentially a species-specific trait, as it is dependent upon longevity and power of flight, there is often an overriding influence of the abiotic and biotic characteristics of the surrounding landscape. In contrast, the number of natural enemies dispersing, and thus their density at a given distance from the release point, is primarily influenced by the number of natural enemies released, and the abundance of the pest relative to the foraging requirements of the released natural enemies. As a result, inoculative releases of natural enemies often pose a much reduced environmental risk in comparison to inundative releases, by virtue of the far smaller numbers of natural enemies released at a site.

Approaches to Quantifying Movement There have been three different approaches used to quantify the dispersal of insects (Fagan, 1997; Turchin, 1998): (i) the analysis of density curves in relation to distance from a release point, (ii) the analysis of fluxes of individuals crossing a boundary and (iii) the analysis of movement paths. The recapture of marked individuals in traps placed at successive distances from a release point, and the subsequent analysis of density–distance curves, were pioneered by Dobzhanzky and Wright (1943) in a study of the movement of Drosophila species. Mark-release-recapture (MRR) has been the most widely used approach in the analysis of dispersal and has been applied to insects of all sizes. The second approach is based on observations of the cumulative

count or flux of marked individuals caught at a delimited boundary surrounding a central release point. Fagan (1997) used this technique to determine the dispersal rate of mantids as they moved out from a central point and were caught on tanglefoot bands at the perimeter of square plots. Although this is an interesting alternative approach, it has yet to be used more extensively and is more complex, but may be particularly well suited to the measurement of biases in dispersal (Turchin, 1998). The third approach is to record the movement of individual insects, to map their paths, and to use temporal and spatial coordinates to estimate dispersal rates. Although this is a powerful approach (Turchin, 1998), and has been used for a number of larger insects, particularly butterflies (e.g. Turchin et al., 1991), it is not suitable for monitoring dispersal of small insects and is better applied to investigations of the effects of environmental heterogeneity on movement. Thus, the methodology that is best suited for the assessment of dispersal as an environmental risk of mass releases of invertebrate natural enemies is the analysis of MRR experiments.

General considerations for mark–release– recapture (MRR) experiments In estimating the pattern of dispersal of natural enemies from a central release point through time in MRR studies, there are several key issues that need to be considered: ● It is essential to be able to distinguish the dispersing natural enemies from individuals in the wild population. This is not a problem if the natural enemy is an exotic species without locally established populations or, if as happens in some cases, that wild populations of an indigenous species are absent in the target region. In many cases, however, indigenous natural enemies will have wild-type counterparts in the field, and distinguishing between mass-released and wild individuals will be a major concern requiring use of some form of natural or applied marker.

Methods for Monitoring the Dispersal of Natural Enemies

● Dispersal from a central point is subject to an area-dilution effect, whereby as individuals move further away from a central release point they are spread over a progressively greater area and consequently become more difficult to recapture (Turchin, 1998). To improve the accuracy of recapture data a greater number of recapture points some distance from the release point could be used, although it is often impractical to monitor a greater number of traps effectively. Alternatively, baits, in the form of foods, hosts, or kairomones, have been used to increase the attractiveness of recapture points, and provide a more practical approach to counteracting the dilution effect. Some caveats, however, are that baits often have an unknown sphere of attraction, which may affect the dispersal of individuals in unknown

117

ways, and can lead to interference if their range of attraction is greater than the distance between recapture points. In addition, recapturing too great a number of individuals before they have completed their dispersal can itself bias the dispersal process, posing a dilemma in terms of the trade-off between sampling efficiency and bias (Yamamura et al., 2003). ● Time is an important variable in any MRR experiment, and its influence on the analysis of dispersal data has often been underestimated. Dispersal is a continuous process and the pattern of recaptures in relation to distance from a central release point changes dramatically with time (Fig. 7.1). Some recapture techniques, such as sweep netting or traps deployed for very short time intervals ( δ ). The alternative hypothesis is the hypothesis of equivalence, or H1:|D| = δ. Again, this kind of analysis depends on the knowledge of what a large (biologically meaningful) effect is, and the determination of delta is similarly as difficult as determination of the effect size, as discussed above. Given the large uncertainty in this area, it is difficult to give advice on this, though the general idea is appealing for decision-makers in risk assessment (Peterman, 1990). In conclusion, a priori PA can be a valuable aid in the design of any study and, in particular, for monitoring programmes (see Barratt et al., Chapter 10, this volume). In addition to the information on sample size necessary to detect a given effect, it is also very valuable for reducing the cost of largescale programmes as far as possible. Depending on the research question, post hoc PA also can be very useful, particularly because it is not always possible to conduct an ideally high number of replicates. It should be stressed that it is not possible with PA to associate an unambiguous probability of being correct in not rejecting the null hypothesis although, unfortunately, this has been done quite often in the past (see Peterman, 1990). Instead, it is only possible to argue that, with a probability of (1⫺β), there is no difference from the H0 greater than the effect size. If both the ES and β are small (and consequently the power is high), it is reasonable to conclude that the effect is negligible. It is particularly important in studies on non-target effects that a conclusion from a non-significant statistical result should be subject to the same stringent probability standards as a positive conclusion from a significant statistical result. Power analysis could be used to provide these standards.

Programs available A comprehensive review on this topic was written by Thomas and Krebs (1997), and we do not attempt to provide a similar

Statistical Tools to Improve the Quality of Experiments

229

detailed compilation here. Instead, we would like to refer to some published information – also on the internet – and highlight a few recent developments. Since the influential paper by Thomas and Krebs (1997), some significant advances have been made, wherein some programs are able to calculate the power for regressions, comparisons of means (ANOVA and General Linear Models) or proportions (χ 2 tests), for correlation tests and survival analysis. However, there are still several statistical tests for which PA is not available and, unfortunately, this includes the Generalized Linear Models, which can be a very powerful statistical tool for data that do not follow a Gaussian distribution. There are also possibilities for calculating power for other tests, but efforts to do this can vary from relatively simple to challenging. For instance, Monte Carlo simulations can be used to calculate power for non-parametric tests (Peterman, 1990). Alternatively, data have to be transformed to fit the assumptions of tests that allow PA, e.g. log-transformation or squareroot transformation for count data, arc sine square-root transformation for proportions (see, e.g. Quinn and Keough, 2002, or another standard statistics textbook, for further information). Information on programs and their strengths and weaknesses can be also obtained from the following homepages: List of programs (from 1996) (http://www.insp.mx/dinf/stat_list.html) and paper by Thomas and Krebs (1997), (http://www.zoology.ubc.ca/~krebs/power. html).

able to detect a 20% loss of biodiversity (i.e. 6.4 species on average), the resulting transformed means for species numbers would be 3, 3, and 2.72 for the three treatments, respectively, and the standard deviation would be approximately 0.5 for all treatments. A simple ANOVA did not detect a significant effect. Entering the above-mentioned values in a programme for PA returns an effect size of ES = 0.2828, and thus what is conventionally described as medium effect size. With a total sample size of 30 the power is (1⫺β) = 0.2397. How many replicates would be needed to achieve a power of 0.8 with such an effect size? Using an a priori test in the programme for PA we receive a necessary sample size of n = 126. Thus, to demonstrate with high confidence that no effect exists would require a much larger study (see, e.g. Lang, 2004 for an estimate of necessary sample sizes for non-target effects of Bt-plants). Using another example, let us see how large the sample size should be in a nontarget effects study of an insect natural enemy. Using the above-mentioned example of Thomas (1997), where the non-target population would be affected only if the mortality were higher than 40%, we can use 0.4 as effect size in an a priori test. If we were to achieve a power of 0.8, the necessary sample size in an experiment with two treatments would be n = 52.

Examples

In a seminal paper, Hurlbert (1984) published a review with respect to proper replication of 176 field experiments covering 156 papers published in ecological journals between 1960 and 1983. Disturbingly, he found that of the 101 studies applying inferential statistics, 48% contained pseudoreplication. Pseudoreplication occurs whenever ‘inferential statistics are used to test for treatment effects with data from experiments where either treatments are not replicated (though samples may be) or replicates are not statistically independent’ (Hurlbert,

Let us, again, take a look at the example data provided in the introduction. Using ten fields for each treatment, the effect of GM plants on insect biodiversity was tested. If we were to analyse those data with ANOVA, we would have to transform the species numbers to receive data with Gaussian distribution. Square root transformation (y ⬘= √៮៮៮៮៮ y+1) could be favourable in our case. If our control plots could harbour eight non-target species and we wish to be

Avoid Being Trapped in Pseudoreplication

230

T.S. Hoffmeister et al.

1984). Statistical independence means that each individual data point might positively or negatively deviate from the population average due to random variation not related to the deviation of another point. An example of lack of statistical independence is given in the introduction, where samples of a study on effects of GMOs on biodiversity were segregated by treatment and, thus, differences attributed towards the treatment could equally well have been attributed to some factor typical for the section of the field the samples came from. In this case, the effects of treatments are potentially confounded with inherent differences between field plots. Although the awareness of researchers of avoiding pseudoreplication has increased and fewer studies contain analyses with pseudoreplicated samples, Heffner et al. (1996) and Ramirez et al. (2000) found, in a recent study on pseudoreplication in experiments on the olfactory response of insects, that an alarming 46% of 105 studies were pseudoreplicated, because of either a lack of independence in the stimulus or the experimental device, the repeated use of experimental animals or the use of groups of animals. Thus, pseudoreplication is still an issue in the design of experiments, and much care has to be taken to avoid any spatial or temporal segregation of samples from different treatments. For example, when testing the host specificity of biological control agents, it is essential that insects for the tests on non-target hosts do not come from one rearing container or incubator and control animals (for the test on target hosts) come from another, or that non-target hosts are always tested in the same container or field cage or on the same plant while target hosts are tested in another cage or on another plant. Equally, positions of experimental units within an experimental chamber or on a field plot need to be switched between treatments to avoid confounding effects of differences in temperature and light conditions, etc. In the same manner, the full set of trials on non-target hosts should not be conducted before tests with target hosts are carried out. Randomization of testing order, or random assignment to

plants or test cages, ensures that pseudoreplication can be avoided. For further reading, we encourage the reader to take a look at the section on pseudoreplication in Ruxton and Colegrave (2003).

Experimental Design: is Randomization Feasible? Basic textbooks on statistics always stress the point that, in order to draw relevant conclusions from an experiment, all treatments, replicates, etc., should be randomized. But what does that mean? Randomization is a process that assigns each replicate of each measured unit (animal, field, species, etc.) to each treatment in a random order, rather than by choice. By doing this, any effect observed will be unequivocally attributed to the treatment studied, and not to lurking variables or uncontrolled factors which might vary over the length of the experiment. For example, if one was interested in estimating the host-range specificity of different potential biological control agents for a pre-release evaluation of non-target risks, he/she would sequentially offer several potential host species to the different biological control agents studied (see van Lenteren et al., Chapter 3, this volume for a detailed description of the proposed method to be used). In this case, it would be preferable to: (i) test the different host species in a random order for the different biological control agents, and (ii) test each host species, with the different biological control agents taken in random order as well. Indeed, in the case where the different host species are always tested in the same order, uncontrolled factors varying with the duration of the experiment could influence the results and lead to differences that might be wrongly interpreted as being due to differences between species. Also, if all potential host species are tested successively on each biological control species, a difference observed between biological control species might simply be due to uncontrolled factors varying with the total duration of the experiment.


The goal of randomization is to produce comparable groups of replicates in terms of general animal, field, etc., characteristics and other key factors that might affect the outcome of the result obtained. In this way, all groups of replicates are as similar as possible at the start of the study. At the end of the study, if group outcomes differ between each other, the investigators can conclude with some confidence that the treatment tested really influenced the results obtained. Most of the time, randomization is performed by means of a computer program, coin flips or a table of random numbers to assign each measured unit to a particular treatment. Advanced additional methods are sometimes used. Is randomization always feasible, especially in evaluating non-target risk in biological control programmes? Unfortunately, the answer is likely to be ‘no’. In the example given above, where we wanted to estimate the host-range specificity of different potential biological control agents, it would probably be unrealistic to design an experiment in which all host species tested and all potential biological control agents compared were randomized. Regarding the fact that the experimental scheme is based on a succession of different measures (see van Lenteren et al., Chapter 3, this volume), having everything randomized would indeed imply having available, during the total duration of the experiment, a sufficient number of all host and biological control agent species at the right stage. In most cases this would simply be not feasible for economic or spatial reasons. All of this should be kept in mind and, if real randomization appears not feasible, results of the experiments should thus be interpreted with caution.

A Unified Approach Instead of a Menu of Tests, General and Generalized Linear Models When the traits to be analysed follow a Gaussian (also called ‘Normal’) distribution, standard t-tests, ANOVA or regression

231

analyses can be used to statistically test the effect of a treatment. All these different ‘classical’ methods assume that the distribution of residuals around the fitted model (i.e. the error distribution) is normal (Gaussian). These different methods, which most readers will be familiar with, are called ‘General Linear Models’, since in its simplest form, a linear model specifies the (linear) relationship between the variable (or response) y, to be explained (the socalled ‘dependent’ variable), and a set of predictors, independent variables, the xs, such that E(y) = b0 + b1x1 + b2x2 + … + bkxk

(1)

In this equation, b0 is the regression coefficient for the intercept and the bi values are the regression coefficients (for variables x1 to xk) computed from the data. So, for example, one could estimate (i.e. predict) the weight of a parasitoid female as a function of the type and number of hosts it feeds on. For many data analysis problems, estimates of the linear relationships between variables are adequate to describe the observed data, and to make reasonable predictions for new observations. However, as we have seen previously (see Box 13.1), most of the biological traits that have to be measured to estimate non-target risks of biological control agents do not necessarily follow a Gaussian distribution. In such cases, the relationship between the variable (or response) y to be explained cannot adequately be summarized by a simple linear equation, for two major reasons: OF THE DEPENDENT VARIABLE. First, the dependent variable of interest may have a non-continuous distribution and, thus, the predicted values of the statistical model should also follow the respective distribution. Any other predicted values are not logically possible. For example, an investigator may be interested in predicting one of two possible discrete outcomes (e.g. a host is accepted or not). In that case, the dependent variable can take on only two distinct values, and the distribution of the dependent variable is said to be binomial. Another example would be to predict how

DISTRIBUTION

232


Box 13.1. Measurement variables and their distribution Many ‘classical’ statistical approaches rely upon the assumption that the probability distribution of data from samples and the error terms of the statistical analyses (the residuals) are distributed normally, i.e. Gaussian. With many of the measurement variables we collect in non-target testing of biological control agents, these assumptions are not met. Count data such as, e.g. number of mature eggs of a female, are usually Poisson distributed, data for percentages are Binomial, and data for longevity are usually Exponential or sometimes Gamma distributed. In theory, it is possible to transform many kinds of data such that the assumptions of parametric tests are met, and those tests are also robust against small deviations from the assumptions; but first of all it is hard to estimate the extent of the robustness against deviations from normality in error terms and, secondly, it is often advisable to use actual data rather than transformed data to meet assumptions. The probable most commonly collected types of data are listed in the table below. Note that deviations from the distributions mentioned in the table might occur in individual cases and that, in general in statistical testing, residuals should always be inspected for the adequacy of the model. Measurement variables often found in non-target testing of biological control agents and their distribution. Measurement variable

Distribution (most likely)

attack rate (per unit time) dispersal capacity diurnal periodicity egg load encounter rate (per unit time) fecundity frequency of mating growth rate host acceptance insertion/deletion of genes latency to attack morphology rate of development rate of predation/parasitism spatial distribution (i.e. counts) survivorship/mortality thermal budget (degree-days)

Gaussian Gaussian, or Poisson if counts Gaussian Poisson if counts or Binomial if proportion Gaussian Poisson if counts or Binomial if proportion Poisson if counts or Binomial if proportion Gaussian Binomial Poisson Gamma Gaussian Gaussian Binomial Poisson or Negative binomial Gamma Gaussian

many females a male can mate with. If we were to study actual numbers and not average number of matings per male, the dependent variable (i.e. number of females mated) is discrete (i.e. a male can mate with one, two or three females and so on, but cannot mate with 3.46 females or with fewer than 0 females), and most likely the distribution of that variable is highly skewed (i.e. most males will mate with one, two or three females, fewer will mate with four or five, very few will mate with six or seven, and so on). In this case it would be reasonable to assume that the dependent variable follows a so-called Poisson distribution.

A second reason why a simple linear model might be inadequate to describe a particular relationship is that the effect of the predictors on the dependent variable may not be linear in nature. For example, the relationship between the fecundity of a synovigenic parasitoid female and its age is most likely not linear in nature. Under standardized conditions, fecundity will not markedly differ between females of one or two days of age, whereas such a difference will probably be greater between older females, even with only one day’s age difference. Probably some kind of a power function would be adequate to

LINK FUNCTION.


describe the relationship between females’ age and fecundity, so that each increment in days of age at older ages will have greater impact on females’ fecundity, as compared to each increment in days of age during early adult life. Put in other words, the link between age and fecundity is best described as non-linear, or rather as a power relationship in this particular example. Generalized Linear Models are a generalization of general linear models and can be used to predict responses both for dependent variables that are not normally distributed and for dependent variables which are non-linearly related to the predictors. Actually, general linear models can be considered as special cases of the generalized linear models. In general, in linear models, the dependent variable values have a normal distribution and the link function, which ‘connects’ the dependent variable to a linear combination of predictor variables, is a simple identity function (i.e. the linear combination of values for the predictor variables is not transformed). To illustrate this, equation (1) gave the general linear model linearly associating a response variable y with values on the x variables, while the relationship in the generalized linear model is assumed to be E(y) = g(b0 + b1x1 + b2x2 + … + bkxk)

(2)

where g(…) is a function. Formally, the inverse function of g(…), say f(…), is called the link function, so that f(E(y)) = b0 + b1x1 + b2x2 + … + bkxk

Various link functions (see McCullagh and Nelder, 1989) can be chosen, depending on the assumed distribution of the y variable values. Table 13.2 gives the four main Generalized Linear Models that can be used in experiments performed to estimate non-target risks of biological control agents. The values of the regression parameters (and their variance and covariance) in the Generalized Linear Model are obtained by a so-called maximum likelihood estimation, which requires iterative computational procedures. Several statistics packages are currently available for doing this. Then, tests of the significance of the effects in the model can be performed via the Wald statistic, the likelihood ratio or score statistic. Detailed descriptions of these tests can be found in McCullagh and Nelder (1989). In summary, Generalized Linear Models are powerful and efficient tools for analysing the sort of data collected in experiments performed to estimate nontarget risks of biological control agents. Just a brief overview has been provided here, and there are several textbooks that provide a thorough description of this sort of statistical modelling approach (e.g. Hosmer and Lemeshow, 1989; McCullagh and Nelder, 1989). We strongly recommend readers of this chapter to consult them.

Examples

(3)

where E(y) stands for the expected value of y.

233

Using again our example from the introduction, we may analyse one of our computer-

Table 13.2. List of the main Generalized Linear Models that can be used in experiments performed to estimate non-target risk of biological control agents. Link functions indicated are the most ‘popular’ ones. Others can be used in particular cases (see McCullagh and Nelder, 1989 for an exhaustive description). Distribution

Model description

Appropriate link function

Type of data analysed

Normal

Traditional linear model

identity: ƒ(y) = y

Normally distributed traits

Binomial

Logistic regression

logit: ƒ(y) = log{y/(1⫺y)}

Fractions (proportions)

Poisson

Log-linear model

log: ƒ(y) = log(y)

Counts

Gamma

Gamma model with inverse link

inverse: ƒ(y) = 1/y

Time durations

234


generated data sets using a Generalized Linear Model. Since we count the number of species in each field plot, our data are most likely Poisson distributed. Specifying a Generalized Linear Model with Poisson distribution and log link function, and using the number of species per plot as response variable and the crop treatment (GM-plants, non-GM isoline and conventional crop) as factor, we find a P-value of 0.0962; thus, there is an insignificant trend in the data (Fig. 13.2a). An analysis of these data using an ANOVA on square root-transformed data yields a P-value of 0.147. A visual comparison (Fig. 13.2b and c) and statistical tests of the normality of the standardized residuals from both analyses (P = 0.515 and P = 0.474, respectively) suggest that the Generalized Linear Model is the slightly more adequate approach to analyse these data. Note that in both cases the statistical result is insignificant and, thus, the null hypothesis of no effect cannot be rejected, but also that the power analysis suggests a lack of power to conclude with confidence that there is no effect. As a second example, imagine a large arena choice test as suggested by van Lenteren et al. (Chapter 3, this volume). We have three different treatments, with ten field cages each: (1) with the target host (or prey, which is used synonymously here) and non-target host present in the same field cage together with the natural enemy,

(2) with only the non-target host and the natural enemy in the same field cage, and (3) with only the target host and the natural enemy in the same field cage. We are interested in whether the target host is killed at a higher rate than the non-target host and whether the mortality of the non-target host depends upon the fact of whether the target host is available to the natural enemy or not. We will not test whether the mortality rates of target and non-target host are equal within treatment (1), because these data would not be independent. Rather, we will test whether the mortality of non-target hosts in treatment (1) is equal to the mortality of non-target hosts in treatment (2) and equal to that of the target hosts in treatment (3) (this is our null hypothesis). Again, we will use computer-generated data. Given that the mortality rates found were 4.1%, 10.6% and 50.5% in (1), (2) and (3), respectively, we use a Generalized Linear Model with binomial distribution and logit link and find a significant effect overall and also between treatments (Table 13.3). Thus, in this example, the non-target host is attacked at a relatively low rate, and even less so when target hosts are available. This result is visible from the estimates in Table 13.3, where the estimate for mortality is positive and thus higher in treatment (2) than in treatment (1), and much higher (more than three times higher) in treatment (3) than in treatment (1).

Fig. 13.2. Simulated average (+ SE) effect of plant treatment on non-target insect species (panel a). The computer-generated data were analysed by means of a Generalized Linear Model with a log link function and ANOVA on square root transformed values, respectively. Panels (b) and (c) show Normality Plots for the standardized residuals of the respective tests. The relationship in panel (b) shows a slightly better fit with normality assumptions than in panel (c).


235

Table 13.3. Results of a Generalized Linear Model on computer-generated data for the mortality rates of target and non-target hosts in large arena choice tests, using an experimental set-up as suggested by van Lenteren et al. (Chapter 3, this volume) (for details, see text). Parameter

Treatment

Estimate

DF

χ-Square

Pr > ChiSq

(3) (2) (1)

⫺3.2591 3.2511 1.0839 0

1 1 1 0

378.47 329.63 30.14 0.0000

ChiSq 0.0025

Pr > Z

Environmental Impact of Invertebrates For Biological Control 0f Anthropods: Methods and Risk Assessment

Methods of Environmental Impact Assessment

Coastal and Estuarine Risk Assessment (Environmental and Ecological Risk Assessment)

Environmental Contaminants: Assessment and Control

Environmental Contaminants: Assessment and Control

Handbook of Environmental Risk Assessment and Management

Environmental Impact Assessment: Theory and Practice

Radiological Risk Assessment and Environmental Analysis

Radiological Risk Assessment and Environmental Analysis

Environmental Risk Assessment of Genetically Modified Organisms

Environmental and Agricultural Modeling: Integrated Approaches for Policy Impact Assessment

Environmental and Economic Sustainability (Environmental and Ecological Risk Assessment)

Environmental Impact Assessment Methodologies, 2nd edition

Tolerance to Environmental Contaminants (Environmental and Ecological Risk Assessment)

Risk Assessment and Risk Management

Handbook of Biological Control: Principles and Applications of Biological Control

Environmental Assessment of Estuarine Ecosystems: A Case Study (Environmental and Ecological Risk Assessment)

Environmental Assessment of Estuarine Ecosystems: A Case Study (Environmental and Ecological Risk Assessment)

Modern Biogeochemistry: Environmental Risk Assessment, Second Edition

Forensic Assessment of Violence Risk: A Guide for Risk Assessment and Risk Management

Environmental Impact of Textiles

Environmental Impact of Textiles

Environmental toxicology and risk assessment: biomarkers and risk assessment, fifth volume

Environmental and Health Risk Assessment and Management: Principles and Practices

Risk Assessment

Immunotoxicology and Risk Assessment

Reliabilty and Risk Assessment

Toxicokinetics and Risk Assessment

Risk Assessment

Risk Assessment

Risk Assessment

Environmental Impact of Invertebrates For Biological Control 0f Anthropods: Methods and Risk Assessment

Methods of Environmental Impact Assessment

Coastal and Estuarine Risk Assessment (Environmental and Ecological Risk Assessment)

Environmental Contaminants: Assessment and Control

Environmental Contaminants: Assessment and Control

Handbook of Environmental Risk Assessment and Management

Environmental Impact Assessment: Theory and Practice

Radiological Risk Assessment and Environmental Analysis

Radiological Risk Assessment and Environmental Analysis

Environmental Risk Assessment of Genetically Modified Organisms

Environmental and Agricultural Modeling: Integrated Approaches for Policy Impact Assessment

Environmental and Economic Sustainability (Environmental and Ecological Risk Assessment)

Environmental Impact Assessment Methodologies, 2nd edition

Tolerance to Environmental Contaminants (Environmental and Ecological Risk Assessment)

Risk Assessment and Risk Management

Handbook of Biological Control: Principles and Applications of Biological Control

Environmental Assessment of Estuarine Ecosystems: A Case Study (Environmental and Ecological Risk Assessment)

Environmental Assessment of Estuarine Ecosystems: A Case Study (Environmental and Ecological Risk Assessment)

Modern Biogeochemistry: Environmental Risk Assessment, Second Edition

Forensic Assessment of Violence Risk: A Guide for Risk Assessment and Risk Management

Environmental Impact of Textiles

Environmental Impact of Textiles

Environmental toxicology and risk assessment: biomarkers and risk assessment, fifth volume

Environmental and Health Risk Assessment and Management: Principles and Practices

Risk Assessment

Immunotoxicology and Risk Assessment

Reliabilty and Risk Assessment

Toxicokinetics and Risk Assessment

Risk Assessment

Risk Assessment

Risk Assessment

Recommend Documents